Using a Radical Scavenger and Modeling the
Improvement of Linewidth in One-Photon
Direct Laser Writing Lithography
A Senior Project submitted to
The Division of Science, Mathematics, and Computing
of
Bard College
by
Shuyi Weng
翁舒逸
Annandale-on-Hudson, New York
May, 2014
Dedication
To My Mom
致母亲
i
Acknowledgements
This project would not have been completed without the help and support from my
professors, friends, and family members. I would like to offer my special thanks to my
project advisors, Dr. Christopher LaFratta and Dr. Csilla Szabo, for their helpful sug-
gestions, careful supervision, and inspiring guidance in both realms of chemistry and
mathematics. I am particularly grateful for the assistance given by my academic advi-
sor, Dr. Emily McLaughlin, who inspired me with the idea of radical scavenging with
knowledge from organic chemistry. I would also like to thank Dr. Jennie D’Ambroise, for
her suggestion of using segmented PDEs to approach the modeling. I wish to acknoledge
the assistance provided by my lab peers Olja Simoska, Ian Pelse, Miles Ingram, Zhe Mei,
Min Kyung Shinn, Jasper Williams, Ingrid Stolt, and Clare Wheeler, as well as all the
other chemistry seniors. The constant support from my friends Seo Young Kim, Yan Chu,
Yushan Jiang, Zexi Song, Jin Zhang, Xujun Liu, and Thinh Pham, has also been a great
pushing force for my work. Last but not least, I would like to thank my family members
for their care and suppport on me, even though hardships have struck us harshly in the
past year.
ii
Abstract
Direct laser writing (DLW) lithography is a useful way to generate microscopic patterns
for lab-on-a-chip devices. We have developed a simple and inexpensive platform for the
creation of these initial patterns using a homemade acrylic resin, a 405-nm laser diode,
and a fluorescence microscope. To create sub-micrometer linewidths, we included a radical
scavenger in the photoresist to quench excessive amounts of radicals generated from laser
exposure. In this project, we compared various formulations of the photoresist and we
presented a model for the radical polymerization in our photoresist. The model utilized the
method of finite differences to numerically solve a system of parabolic partial differential
equations describing the diffusive behaviors of particles in the photoresist. The model also
made use of Markov chains to simulate the step-wise polymerization of the photoresist.
The parameters in the model were determined by fitting the model on experimentally-
obtained data.
iii
Contents
Dedication i
Acknowledgements ii
Abstract iii
I Laboratory Experiments
1 Introduction I 2
1.1 Photolithography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Direct Laser Writing (DLW) . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Acrylic Ester Polymerization . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Memory Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Radical Inhibition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Experimental Overview 12
2.1 Laser Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Microscope and Automated Stage . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Photoresist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 Modified Glass Slides or Coverslips . . . . . . . . . . . . . . . . . . . . . . 19
2.5 Constructing a Micro-Scale Structure . . . . . . . . . . . . . . . . . . . . . 19
iv
CONTENTS v
2.6 Developing a Micro-Scale Structure . . . . . . . . . . . . . . . . . . . . . . 22
2.7 Scanning Electron Microscopy (SEM) . . . . . . . . . . . . . . . . . . . . . 22
3 Shrinkage Study 23
3.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4 Linewidth Characterization 29
4.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
II Mathematical Modeling
5 Introduction II 36
5.1 The Gaussian Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Parabolic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.3 Method of Finite Differences . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.4 Markov Chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6 Memory Effect Characterization 48
6.1 Oscilloscope Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7 Modeling the Memory Effect 54
7.1 First Approximation: Radical Initiation . . . . . . . . . . . . . . . . . . . . 56
7.2 Second Approximation: Acrylate Initiation . . . . . . . . . . . . . . . . . . 59
CONTENTS vi
7.3 Probability of Polymerization . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.4 Third Approximation: Markov Modeling . . . . . . . . . . . . . . . . . . . 63
8 Mathematica Implementation 65
8.1 Finite Difference Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
8.2 Markov Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
8.3 Pseudocode for the Implementation . . . . . . . . . . . . . . . . . . . . . . 67
8.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.5 Future Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
A Linewidths before/after Alcohol Development 75
B Linewidth Data with Different DLW Conditions 82
C SEM Images of Double Dots 88
D Coding in Mathematica 93
D.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
D.2 Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
D.3 First Exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
D.4 Second Exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
D.5 Showing the Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
E Sample Simulation 100
F General Solutions to y
0
= ky(1 − y) 104
List of Figures
1.1 Photoinitiator Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Laser Beam Going through the Photoresist . . . . . . . . . . . . . . . . . . 7
1.3 Acrylic Acid and Acrylic Ester . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Acrylic Ester Polymerization . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Example of “Memory Effect” . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 Inhibition Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Laser Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Comparison between Airy Disk and Gaussian Function . . . . . . . . . . . 15
2.3 Microscope Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Direct Laser Writing Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Photoresist Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6 Hybrid Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1 Shrinkage Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Structure Used to Measure the Shrinkage Coefficient . . . . . . . . . . . . 25
3.3 Waffle Structure and Measuring Scheme . . . . . . . . . . . . . . . . . . . 27
4.1 Straight Lines for Linewidth Measurements . . . . . . . . . . . . . . . . . . 30
4.2 Linewidth of DLW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.3 Linewidths under 150 nW, 20µm/s (after shrinkage) . . . . . . . . . . . . . 34
vii
LIST OF FIGURES viii
4.4 Linewidths under 150 nW, 20µm/s (before shrinkage) . . . . . . . . . . . . 34
5.1 Beam Radius as a Function of z . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Difference Grid on the Space-Time Domain . . . . . . . . . . . . . . . . . . 44
5.3 A State Diagram of the Markov Chain for Directional Random Walk . . . 46
6.1 Digital oscilloscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2 Merged Double-Dot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7.1 Beam Intensity and Reflection Intensity . . . . . . . . . . . . . . . . . . . . 58
7.2 Modeling the Probability Function . . . . . . . . . . . . . . . . . . . . . . 62
8.1 Simulation Result of Experiment #2 . . . . . . . . . . . . . . . . . . . . . 71
8.2 Final State of Simulation #8 . . . . . . . . . . . . . . . . . . . . . . . . . . 72
List of Tables
2.1 Composition of Different Photoresists . . . . . . . . . . . . . . . . . . . . . 18
3.1 Percent Shrinkage of Photoresists . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Shrinkage in Microscale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.1 Laser Powers and Writing Speeds for Linewidth Characterization . . . . . 30
4.2 Calculated Linewidths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.1 Base Speed Range for Specific Hiatus Time . . . . . . . . . . . . . . . . . . 50
6.2 Memory Effect Measurements . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.1 Table of Variables and Coefficients . . . . . . . . . . . . . . . . . . . . . . 55
7.2 Table of Functions and Lists . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.1 Choices for Parameters and Functions . . . . . . . . . . . . . . . . . . . . . 69
8.2 Percent Increase Results Simulated by the Model . . . . . . . . . . . . . . 70
8.3 Comparison between Experiments and the Model . . . . . . . . . . . . . . 72
A.1 Shrinkage Characterization for Photoresist #1 . . . . . . . . . . . . . . . . 76
A.2 Shrinkage Characterization for Photoresist #2 . . . . . . . . . . . . . . . . 76
A.3 Shrinkage Characterization for Photoresist #3 . . . . . . . . . . . . . . . . 77
A.4 Shrinkage Characterization for Photoresist #4 . . . . . . . . . . . . . . . . 77
B.1 Linewidth Measurement for Photoresist #1 . . . . . . . . . . . . . . . . . . 83
ix
LIST OF TABLES x
B.2 Linewidth Measurement for Photoresist #2 . . . . . . . . . . . . . . . . . . 84
B.3 Linewidth Measurement for Photoresist #3 . . . . . . . . . . . . . . . . . . 85
B.4 Linewidth Measurement for Photoresist #4 . . . . . . . . . . . . . . . . . . 86
B.5 Linewidth under Different Condition . . . . . . . . . . . . . . . . . . . . . 87
Part I
Laboratory Experiments
Chapter 1
Introduction I
The rapid development of science and technology has demanded the construction of
smaller and smaller structures for various reasons. In the realm of computer engineering,
Moore’s Law [Moo65], proposed in 1965, accurately predicted that the number of tran-
sistors on an integrated circuit would double approximately every two years. In order for
millions and even billions of transistors to be created on a single chip, their size has to be
miniature. The shrinking of electronics components and their integration into complex
circuits have enabled the creation of modern electronics and the age of information.
A similar miniaturization revolution is underway in the physical sciences. First pre-
dicted by the physicist Richard Feynman in his famous “There’s plenty of room at the
bottom” lecture, Feynman explains the basic tenets of nanotechnology. The fields of
chemistry and biology are joining physics and the electronics industry in trying to make
technologies smaller, faster, and cheaper. For example, the assays on micro-biological
2
CHAPTER 1. INTRODUCTION I 3
structures require testing chambers of the size comparable to a single bacterium or a
red blood cell [ZNWS13, TBP
+
11], which is usually in micrometer scale. The demands
on micro-structures from biological sciences have given rise to the development of micro
total analysis systems (µTAS) [KOM
+
13], which integrate the functionalities of multiple
chemical assays into one single chip. Such devices are known as “lab-on-a-chip” (LOC)
for their capability to perform various chemical tests and their ability to operate assays
using only picoliters of sample.
The use of micro-scale devices is constantly increasing in chemistry and biology, and so
the study of new fabrication techniques is an ever evolving and important field. Nowadays
in electronics industry, photolithography has become one of the most commonly-applied
techniques for microfabrication [Jae02]. It uses a mask to expose the surface of a layer
of photoresist selectively by ultraviolet (UV) light, chemically changing the photoresist
such that the exposed area is more or less soluble than the unexposed area. After devel-
opment in a solvent, a two-dimensional pattern is left on the substrate. With multiple
runs of such processes, a complicated circuit system can be “drawn” on the substrate.
Besides photolithography, other techniques for the fabrication of micro-structures include
electron beam lithography [RC97], dynamic masking [IKT
+
08], high precision machining
[ZRM
+
03], and direct laser writing. One of the main distinctions between these mothods
and conventional mask-based lithography is their serial nature compared to the parallel
nature of masks.
CHAPTER 1. INTRODUCTION I 4
1.1 Photolithography
Conventional mask-based photolithography is widely practiced in contemporary elec-
tronics industry. A series of sophisticated manipulations are performed under highly-
regulated conditions [Jae02] to produce integrated circuits that go into commercial com-
puter chips and processors.
The whole process of photolithography takes place in a clean room. It transfers pat-
terns from masks to photoresist on the surface of silicon wafers. Commertial chips are
mostly produced on silicon wafers with positive photoresists, in which case the area ex-
posed under UV light turns soluable and can be washed off by solvents. In the subsequent
etching or doping steps, the photoresist protects portions of the surface while windows
are etched in barrier layer such as silicon dioxide, silicon nitride, or metal. After etch-
ing, the photoresist is removed, and impurities can be introduced into the wafer through
the windows, or metal can be deposited on the surface making contact with the silicon
through the etched windows.
Photolithography requires the use of photomasks at the ultraviolet exposure step.
One photomask gives only a specific pattern on the wafer, while that pattern may contain
millions of features that are all made in parallel. Multiple photomasks are required for
different structures drawn on the substrate, which could be inconvenient and expensive.
Hence, some maskless alternatives have been developed to overcome the drawbacks of
photolithography.
CHAPTER 1. INTRODUCTION I 5
1.2 Direct Laser Writing (DLW)
Direct laser writing (DLW) is an effective way to fabricate micro-structures without
a mask. It uses a moving laser to initiate a change in the solubility of the photoresist to
achieve its goal. Similar with those in photolithography, the photoresists used in DLW
can be either positive or negative. Usually in DLW, a negative photoresist consists of at
least a monomer and a photoinitiator. The photoinitiator absorbs photonic energy from
the laser, breaks apart into radicals, and initiates the polymerization process. Depending
on the type of photonic absorption by the photoinitiator, DLW can be categorized into
two major branches: one-photon DLW and two-photon DLW. The free radicals generated
by the photoinitiator subsequently trigger the polymerization reaction of the monomers,
which will be further discussed in section 1.3.
1.2.1 One-photon DLW
Photoinitiators are usually capable of homolytic cleavage at one or more of its chemi-
cal bonds, resulting in the formation of free radicals (Figure 1.1). When a photoinitiator
molecule absorbs a photon with a specific range of wavelengths, the molecule is excited
from the ground state (S
0
) to an excited electronic state (S
∗
1
). The molecule then under-
goes non-radiative relaxation into the vibrational ground state of the first excited state
(S
1
). Intersystem crossing (ISC) occurs at this moment to bring the molecule from a
singlet state (S
1
) to a triplet state (T
1
). The molecules in the triplet state are clasically
CHAPTER 1. INTRODUCTION I 6
forbidden to make the transition back to the ground state, which makes the lifetime of
excited molecules long enough to yield free radicals.
S
0
S
0
*
S
1
S
1
*
T
1
ISC
Absorption
R
Figure 1.1: A Jablonski Diagram Showing the Activation of the Photoinitiator
1.2.2 Two-photon DLW
Normally, a molecule absorbs only one photon to make transitions between its elec-
tronic states. However, when two photons hit a molecule simultaneously (which rarely
happens in nature because the photon intensity required is extremenly high), the molecule
is able to make a transition to an excited state with an energy gap corresponding to the
energy sum of these two photons [Abe62]. Such a process is usually achieved with pulsed
lasers, which produces pulses with high intensity. The highly intense laser pulses make
two-photon absorption possible in a tiny volume element, or voxel, at the focal point of the
laser, thus generating a limited amount of free radicals that are only able to polymerize
the photoresist at this small scale.
CHAPTER 1. INTRODUCTION I 7
1.2.3 Comparisons between the two types of DLW
In contemporary research labs, both one-photon DLW and two-photon DLW tech-
niques are practiced for their distinctive advantages over one another.
Two-photon DLW gives structures with smaller linewidths and resolution comparing
to one-photon DLW. When the laser beam is focused on the surface of the substrate
(Figure 1.2), it goes through the photoresist, and the intensity of the beam decreases
proportional to the square of the distance from the focal point. In one-photon DLW, a
large range of beam intensity is capable for photoinitiation and polymerization. Even a low
beam intensity leads to a low concentration of free radicals, which may not reach the level
required for polymerization, but enough to generate oligomers that turns into polymer
more easily and more rapidly than unexposed photoresist. However, in two-photon DLW,
the beam intensity plays an important role in the initiation process. For two photons
to reach one photoinitiator molecule simultaneously, a high intensity of the laser beam
photoresist
objective lens
Figure 1.2: Laser beam going through the photoresist, with the gradient illustrating change in
beam intensity
CHAPTER 1. INTRODUCTION I 8
is required. Such high intensity is only attained near the focal point, while further away
from the focal point, the probability for two photons to reach the same molecule at the
same time is essentially zero. Thus the free radicals are only generated within a short
distance of the focus but not deep into the photoresist. The highly-localized radicals in
two-photon DLW yield smaller structures with higher resolutions than one-photon DLW
does.
However, the cost for a pulsed laser used in two-photon DLW is much higher than that
of a continuous wave laser. A femtosecond titanium:sapphire laser used in two-photon
DLW costs more than $100,000 together with a 532-nm pump laser, while a 405-nm laser
diode module is available at $185 on ThorLabs
r
.
1.3 Acrylic Ester Polymerization
Acrylic esters (Figure 1.3) are the most commonly-used monomers in DLW photore-
sists. They undergo radical polymerization when a photoinitiator is used to generate
radicals. The reaction mechanism is shown in Figure 1.4.
Figure 1.3: Acrylic Acid and Acrylic esters
The introduction of multiple acrylate groups in the R chain of the acrylic ester molecule
leads to branching and cross-linking of polymer chains, thus enhancing the strength and
CHAPTER 1. INTRODUCTION I 9
Figure 1.4: A Mechanism for Acrylic Ester Polymerization
rigidity of the polymer. Substances with multiple acrylate groups are often used in DLW
as the monomer [FW13].
1.4 Memory Effect
As shown in Figure 1.2, the beam intensity decreases as the laser beam goes deeper
into the photoresist. The beam intensity reaches its maximum at the focus, promoting
CHAPTER 1. INTRODUCTION I 10
Figure 1.5: An example of “memory effect”, where the lines are drawn from top to bottom
the radical initiation at that point, but photons are also absorbed beyond the focal point.
The concentration of initiated free radicals beyond the focal point may not be high enough
to propagate the growth of long-chain polymers, but the generated radicals do not die
off immediately, and the oligomer molecules linger. The lifetime of the radicals is long
enough to affect the subsequent exposures nearby, resulting in an increase in linewidth
of the structures (Figure 1.5). Additionally, the photoresist, while quite viscous, does
allow for diffusion of radicals, and the reflection of the laser beam within the photoresist
also leads to higher-order exposure. The diffusion of radicals and smaller oligomers made
unintentionally beyond the focal point affect the resolution of DLW, and thus prohibit
the fabrication of structures as small or as close to each other as we wish. Since it usually
appears to us that the photoresist somehow “remembers” previous laser exposures and
CHAPTER 1. INTRODUCTION I 11
affects nearby structures fabricated a short time later, we have named this phenomenon
the “memory effect”.
1.5 Radical Inhibition
Radical scavengers are commonly used in chemistry in order to remove or deacti-
vate unwanted reaction products in a mixture. For example, butylated hydroxytoluene
(BHT) is usually added to the stock of diethyl ether as an inhibitor [Sig13] to prevent the
formation of explosive diethyl ether hydroperoxide, which reacts via a radical pathway.
Figure 1.6: The Mechanism for Radical Inhibition with BHT
The polymerization of monomers depends on the concentration of free radicals in
the photoresist, and here is where the idea of inhibition comes from. One-photon DLW
creates unwanted free radicals and unpolymerized oligomer molecules beyond the focal
point after each exposure. If radical scavengers are added into the photoresist, they will
be able to quench the free radicals, thus preventing further polymerization involving these
free radicals and improving linewidth and resolution in one-photon DLW.
Chapter 2
Experimental Overview
The fabrication of micro-scale structures is carried out with a systematic procedure.
The procedure starts from modifying glass slides or coverslips. Acrylate groups are at-
tached to the surface of the glass substrate so that the polymerized photoresist may bond
tightly to the glass surface. The photoresist in our experiments consists of three sub-
stances - the monomer, photoinitiator, and the radical scavenger. When a collimated
laser beam is focused onto the photoresist, the photoinitiator is excited and turns into
radicals, initiating the radical polymerization process as shown in Figure 1.4. The laser
beam is filtered and collimated by a series of optics carefully positioned on an optical
table equipped with vibration isolation system. The glass slide with photoresist is fixed
on an automated stage that moves in the plane perpendicular to the laser beam. Polymer
structures are fabricated by moving the stage relative to the laser beam, leaving polymers
attached on the modified glass surface.
12
CHAPTER 2. EXPERIMENTAL OVERVIEW 13
In the following sections, descriptions of each experimental procedure and apparatuses
in use are provided in detail.
2.1 Laser Setup
A laser diode module (CPS405, ThorLabs, USA) is used throughout our experiments.
It produces a 405-nm, 4.5-mW elliptical beam. However, a collimated circular beam is
desired for our purposes. Therefore, two converging lenses and a pinhole are installed to-
gether with a few other optical equipments to modulate the beam, as shown in Figure 2.1.
Blue Laser
405 nm
Half-wave Plate
Polarizing
Beam Splitter
Beam
Block
Lens 1
f
1
= 75 nm
Lens 2
f
2
= 200 nm
Pinhole
d
= 10 Μm
Periscope Unit
Microscope
Figure 2.1: A schematic for the collimation of the laser
The beam coming out of the laser diode is plane polarized. The half-wave plate is
a birefringent crystal that rotates the plane of polarization when the crystal is rotated.
Different polarization of the beam results in different amount of reflection from the po-
larizing beam splitter. The combination of these two pieces of equipment allows us to
efficiently control the power output by rotating the half-wave plate.
CHAPTER 2. EXPERIMENTAL OVERVIEW 14
The lenses in this system are placed in a way such that the principle axes of the two
lenses align with the center of the laser beam, and the distance between them equals
the sum of their focal distance. In this fashion, a telescope that expands and collimates
the beam is created. A pinhole of diameter 10 µm is placed at the common focal point
of the lenses. This pinhole acts as a spatial filter to remove unwanted transverse laser
modes. This arrangement results in an output pattern of an Airy disk [Air35, ST91].
The intensity of the Airy disk is symmetric around its center, and the relation between
the intensity and the distance from the center can be expressed in terms of the Bessel
function:
I(θ) = I
0
2J
1
(ka sin θ)
ka sin θ
2
,
where I
0
is the intensity at the center, k = 2π/λ is the wavenumber, a is the radius of the
pinhole, and J
1
is the Bessel function of the first kind of order one. If we let x = ka sin θ,
and plot I(x)/I
0
, we may discover that it resembles the curve for a Gaussian function
(Figure 2.2). In fact, in Part II of this project, we are going to approximate the laser
beam as a Gaussian beam. All the results of mathematical modeling in this project are
based on this approximation.
Once the laser beam passes through the lenses and the pinhole, it is considered nicely
circular and collimated. The beam is now ready to enter the microscope for DLW. The
periscope unit consists of two mirrors that change the direction of the beam and send it
up into the microscope.
CHAPTER 2. EXPERIMENTAL OVERVIEW 15
0
2
4
6
8
10
0.0
0.2
0.4
0.6
0.8
1.0
Radial Coordinate
Normalized Amplitude
Figure 2.2: Comparison between the intensity of Airy disk and Gaussian function
2.2 Microscope and Automated Stage
The collimated beam is first reflected by a dichroic mirror after it enters the microscope
unit. The dichroic mirror reflects wavelengths less than 420 nm and transmits longer
Objective Lens
Laser Set-up
Cover Slip
Photoresist
Glass Slide
Figure 2.3: A schematic for the collimated beam travelling through the microscope
wavelengths, which will be used to image the sample. The beam then goes through the
objective lens and gets focused on the sample we prepared (see Section 2.5) as shown in
CHAPTER 2. EXPERIMENTAL OVERVIEW 16
Figure 2.3. The most commonly used objective lenses are the 4× objective and the 40×
oil objective, which we will discuss further in Section 2.5. A picture of the complete DLW
setup is taken and shown below in Figure 2.4.
Figure 2.4: Direct Laser Writing Setup (Photo credit: Olja Simoska)
The sample is fixed on an automated stage, which moves in the plane perpendicular
to the principle axis of the laser beam. When the laser beam is turned on, polymerization
of the photoresist occurs at the spot of exposure. If the automated stage is moving at the
same time, a trace of polymerized photoresist shows up on the path in which the focal
point of the laser beam moves relative to the sample. A series of planar structures can
be fabricated by pre-programming the moves of the stage with a graphical user interface,
which will be further discussed in Section 2.5.
CHAPTER 2. EXPERIMENTAL OVERVIEW 17
2.3 Photoresist
The photoresists that we used throughout the experiments consists of an acrylic ester
(monomer), a photoinitiator, and an inhibitor. As described in Sections 1.2, 1.3, and
1.5, the photoinitiator absorbs photonic energy from the laser and turns into radicals.
The radicals propagate to form polymer chains as they react with the carbon-carbon
double bonds in the acrylate groups. The inhibitor in the photoresist prevents “over-
polymerization” by quenching the excessive free radicals generated by photoinitiation.
2.3.1 Composition
The structures of the three components of the photoresist are shown in Figure 2.5.
The majority (≥ 95%) of the photoresist is composed of the acrylate monomer, which,
in our experiments, is dipentaerythritol-pentaacrylate (SR-399) (Sartomer, USA). The
photo-initiator used is Lucirin
r
TPO-L (BASF, Germany). We control the concentration
of the photo-initiator to be 1%. Trial experiments have shown that this concentration
works the best for a wide ranges of exposure dose, as well as both our objective lenses. Dif-
Figure 2.5: Photoresist Composition
CHAPTER 2. EXPERIMENTAL OVERVIEW 18
ferent concentrations (0 ∼ 4%) of radical scavenger 2-tert-butyl-4-methylphenol (tBMP)
(Sigma-Aldrich, USA) were also added into the photoresist as the inhibitor.
2.3.2 Preparation
Four types of photoresists with different concentrations of tBMP are prepared accord-
ing to the mass percentage listed in Table 2.1. Notice that the most significant difference
among them is the concentration of tBMP. In later experiments, we will investigate the
effects of inhibitor on different properties in DLW.
Photoresist # SR-399 Lucirin
r
TPO-L tBMP
1 99% 1% 0%
2 97% 1% 2%
3 96% 1% 3%
4 95% 1% 4%
Table 2.1: Composition (Mass Percent) of Different Photoresists
Because the photo-initiator is photo-sensitive, the preparation of each photoresist is
carried out in a dark room. A light-insulating bottle is made by covering the surface
of a 20-mL glass vial with aluminum foil as the container for the photoresist. After all
compounds are weighed properly and added into the bottle, the mixture is heated to
80 ∼ 90
◦
C. By reaching this temperature range, tBMP melts in the mixture, and the
viscosity of the mixture is reduced dramatically for better mixing. At this time, a wooden
CHAPTER 2. EXPERIMENTAL OVERVIEW 19
applicator is used to mix the components. The photoresist is then cooled in a hood and
is ready for use after it gets back to room temperature.
2.4 Modified Glass Slides or Coverslips
Glass is made from silicate salts, and silicate is usually not a good substrate for
microfabrication because of its low affinity for acrylate polymers. However, if we perform
a modification on the glass surface, the acrylate polymer will attach covalently to the
glass surface, resulting in permanent structures on the glass.
Clean, dry glass slides or coverslips (1 in × 1 in) are cleaned in an air plasma (Harrick
Scientific model PDC-32G) for 5 minutes, and then placed in a solution containing 2%
(3-acryloxypropyl)-trimethoxysilane (Gelest, USA), 3% water, and 95% ethanol (Sigma
Aldrich, USA) for 6 hours with constant stirring. The acrylate-modified coverslips are
then rinsed with ethanol twice, dried with nitrogen gas, and heated in an oven at 100
◦
C
for 20 minutes. At this point, the surface of the glass slides or coverslips is covered with a
layer of acrylate groups. When polymers form under laser exposures, the polymer chains
bond to the acrylate groups, leaving tightly bonded structures on the glass surface.
2.5 Constructing a Micro-Scale Structure
A sample for DLW is prepared by placing a drop of photoresist between a clean, dry
microscope slide (3 in × 1 in) and an acrylate modified glass surface. Depending on the
CHAPTER 2. EXPERIMENTAL OVERVIEW 20
structure we desire to construct, we may use a modified glass slide or coverslip. The
difference between these two types of substrates and fabrication techniques are discussed
in the following subsections.
2.5.1 Fabrication using the 4× objective lens (NA = 0.10)
Both modified glass slides and coverslips work for this setup. The sample for DLW is
fixed on the automated stage above the objective lens. The automated stage, together
with the shutter in front of the laser diode, is controlled by a program (ProScan GUI,
Miles Ingram) on a computer. Desired structures can be programmed and fabricated
automatically with the program. The moving speed of the photonic exposure can be
programmed as the moving speed of the automated stage. The power output of the laser
diode can be controlled with the half-wave plate and the polarizing beam splitter (Section
2.1). An effective power for DLW with 4× objective lens ranges from 1 µW to 500 µW,
depending on the fabrication speed.
2.5.2 Fabrication using the 40× oil objective lens (NA = 1.30)
Only modified coverslips work for this setup. Before one starts to fabricate any struc-
ture with the 40× objective lens, a drop of index matched immersion oil must be applied
to the objective lens. The oil drop then comes into contact with both the lens and the
sample on the side of the coverslip. A glass slide cannot be used with this objective
because the focal distance of the objective lens is smaller than the thickness of the glass
CHAPTER 2. EXPERIMENTAL OVERVIEW 21
slide. Structures with much small dimension can be fabricated with the same program as
the one we used for 4× objective lens. An effective power for DLW with 40× oil objective
lens ranges from 50 nW to 300 nW, depending on both the fabrication speed and the
complexity of the structure.
Figure 2.6: A hybrid structure using both objectives
Hybrid structures can be fabricated by combining the advantages of both lenses. A
typical size for DLW with the 4× objective is 50 ∼ 100 µm, while that with the 40× oil
objective is 1 ∼ 3 µm. The general shape of a structure can be first constructed with the
4× objective, and then the delicate details may be added with the 40× oil objective. An
electron micrograph of a successful example is shown in Figure 2.6.
CHAPTER 2. EXPERIMENTAL OVERVIEW 22
2.6 Developing a Micro-Scale Structure
After the desired structure is successfully fabricated with the pre-programmed com-
mands, the DLW sample is removed from the automated stage. If it just underwent
40× fabrication, the index matched immersion oil should be wiped off the coverslip with
hexanes. The sample is then submerged in ethanol for 10 minutes to dissolve the un-
polymerized photoresist. The substrate is carefully separated with the microscope slide,
and is further rinsed by ethanol and methanol. After drying the DLW sample with gentle
flow of air, the structure is ready to be used for further steps including scanning electron
microscopy (SEM) imaging, microfluidics, and soft lithography.
2.7 Scanning Electron Microscopy (SEM)
Before sending the fabricated sample to the SEM, a 20-nm layer of gold is sputter
coated (Cressington, Model 108 Sputter Coater) to the surface of the sample so that the
surface becomes electrically conductive. The sample is then imaged and measured using
a field emission scanning electron microscope (Tescan MIRA 3).
Chapter 3
Shrinkage Study
We notice that the acrylate polymer shrinks in size during the alcohol-rinsing process.
Such shrinkage may affect the accuracy in further studies on memory effect. We also
notice that the shrinkage of the polymer depends on the concentration of the inhibitor.
As shown in Figure 3.1, the polymer created with the inhibitor-containing photoresist
exhibits more significant shrinkage than the other. Here we present a set of methods to
quantitatively characterize the shrinkage of the acrylate polymer.
Figure 3.1: Comparison between the shrinkage of two photoresists
23
CHAPTER 3. SHRINKAGE STUDY 24
3.1 Methods
A sample for DLW was prepared as described in Section 2.5. With the 4× objective
lens, a set of lines with different linewidths ranging from 50 µm to 200 µm was fabricated
(100 µW, 500 µm/s) on the modified coverslip. The linewidths of the lines were first
measured under the optical microscope before rinsing. After the first round of measure-
ments, the sample was rinsed with alcohols and dried over blowing nitrogen gas. It is then
sputter-coated and sent to the SEM for measurements after alcohol developing. Two sets
of measurements were tabulated and compared to obtain the shrinkage coefficient for the
photoresist. The process was repeated with four different photoresists.
Near the surface of the modified coverslip, the shrinkage behavior is restricted because
the polymer chains are bonded to the acrylate groups on the surface of glass. We introduce
a new method to characterize the shrinkage coefficients of the photoresists. The data
obtained by this method are of more interst to this project because most of the following
experiments are carried out with 40× oil lens, and so is this method.
A sample for DLW was prepared as described in Section 2.5. With the 40× oil objective
lens, four columns of dimension 10 ×10 ×10 µm were fabricated (80 nW, 50 µm/s) layer
by layer with different focal depth. The columns were then connected with two single-
pass DLW at 10 µm above the surface of coverslip, forming a cross in the middle of the
structure. A 10-by-10-µm grid of lines was subsequently drawn on the cross to yield a
waffle-like structure floating in the center. A 3-D presentation of the structure is shown
CHAPTER 3. SHRINKAGE STUDY 25
in Figure 3.2. The sample was then rinsed with alcohols and dried carefully by letting the
alcohols evaporate. As the alcohols evaporated away, the waffle structure was pulled onto
the surface of the glass due to surface tension of the remaining fluid. After the sample
was completely dried, it was sent for sputter-coating and SEM imaging to determine the
dimension of the waffle structure after alcohol developing.
Figure 3.2: Structure Used to Measure the Shrinkage Coefficient
3.2 Results
The results are generally consistent with our conjecture. We present the results from
different methods in two separate subsections.
CHAPTER 3. SHRINKAGE STUDY 26
3.2.1 Shrinkage with the 4× objective lens
The measured linewidths before and after alcohol developing are tabulated in Table A.1
– A.4 in Appendix A. For each photoresist, we plot the linewidth before developing v.s.
the linewidth after (Figure A.1 – A.4). The slope of the plot indicates the proportion of
remaining linewidth, which can be used to calculate percent shrinkage of each photoresist
(Table 3.1).
Photoresist # Regression Formula % Remaining % Shrinkage
1 y = 0.870413x 87% 13%
2 y = 0.876853x 88% 12%
3 y = 0.856793x 86% 14%
4 y = 0.803336x 80% 20%
Table 3.1: Percent Shrinkage of Each Photoresist
3.2.2 Shrinkage with the 40× oil objective lens
The highly concentrated exposure dose for fabricating the structure illustrated in Fig-
ure 3.3a gave rise to troubles in measuring the actual dimension of the waffle structure
because the central structure had increased in size due to memory effect (Section 1.4).
Fortunately, the polymer retained its waffle-like texture so that we were able to measure
the distance between the first and the last swells of the structure in both x and y direc-
CHAPTER 3. SHRINKAGE STUDY 27
tions. SEM images of the fabricated structure and measuring scheme are shown below in
Figure 3.3. The results are taulated in Table 3.2.
(a) Waffle Structure at 45
◦
of View (b) Measuring the waffle structure
Figure 3.3: Waffle Structure and Measuring Scheme
Photoresist # Original Size
After Developing
% Shrinkage
x-direction y-direction
1 10 µm 8.04 µm 8.04 µm 20%
2 10 µm 7.67 µm 7.91 µm 22%
3 10 µm 7.37 µm 7.61 µm 25%
4 10 µm 7.20 µm 7.29 µm 28%
Table 3.2: Shrinkage Coefficients of Photoresists under 40× Oil Condition
CHAPTER 3. SHRINKAGE STUDY 28
3.3 Discussion
The measurements with both 4× and 40× objective lenses showed that the percent
shrinkage increases as the concentration of inhibitor in the photoresist increases. This
conclusion explained the different appearance of polymers under the same fabricating
conditions with different photoresists (Figure 3.1).
The results obtained from the 4× air objective lens appeared to be more controversial.
The percent shrinkage of the first three photoresists are close in range (12% - 14%).
However, there were more systematic errors involved in this method: (1) the shrinking
restriction and (2) the power dependence.
The polymers fabricated with 4× lens bond to the surface of the glass, preventing
it from shrinking near the base. In the 40× case, the polymers to be measured were
fabricated without contacting the surface of the glass, and thus able to shrink evenly in
all directions.
The precent shrinkage also appeared to be power-dependent in the 4× case. For
instance, the separated branches in the plot for photoresist #1 (Figure A.1) were the
outcome of changing the laser power (from 10 µW to 100 µW) during the experiment.
However, the percent shrinkage is less power-dependent in the 40× case because the useful
power range for 40× oil fabrication (50 nW ∼ 300 nW, see Chapter 4) is much narrower
than that for 4× fabrication.
Chapter 4
Linewidth Characterization
One of the most fundamental and most useful structure for DLW lithography is a
straight line. In this chapter, we present a study on the properties of straight lines
fabricated from DLW under different conditions. We used different laser powers, writing
speeds, and inhibitor concentrations, and compared the linewidths under these conditions.
The results gives a table of reference for future fabrication of lines with desired linewidth.
4.1 Methods
A sample for DLW was prepared as described in Section 2.5. A straight line was
first drawn on the coverslip while adjusting the z-coordinate in order to align the focal
point of the laser with the surface of the coverslip. A set of five straight lines with a
specific power-speed combination was drawn 100 µm away from the focus-adjustment line
29
CHAPTER 4. LINEWIDTH CHARACTERIZATION 30
in order to prevent significant memory effects. The sample was developed and sputter-
coated, then sent for SEM imaging as described in Sections 2.6 and 2.7. An SEM image
of the fabricated straight lines is shown in Figure 4.1. Such process was repeated for all
the power-speed combinations for the following choices of laser powers and writing speeds
in Table 4.1. The measured linewidth under each condition is tabulated and compared
for the analysis of the linewidth’s dependence on different conditions.
Figure 4.1: Straight Lines for Linewidth Measurements
Laser Power (nW) 100 150 200 250
Writing Speed (µm/s) 20 30 40 50
Table 4.1: Laser Powers and Writing Speeds for Linewidth Characterization
CHAPTER 4. LINEWIDTH CHARACTERIZATION 31
In order to reduce the unexpected variations during the fabrication, and to save the
usage of modified coverslips, all the sixteen combinations were operated on one sample.
Then the same combinations were repeated for the other photoresists. In the end, we
obtained a total of 64 sets of lines for linewidth measurements.
4.2 Results
The raw data of linewidth measurements on the 64 sets of lines are tabulated in
Appendix B. The “not available” (N/A) data points are conditions under which linear
fabrication was difficult, if not impossible, due to low laser power and/or high writing
speed. The average linewidths obtained from different fabricating conditions are tabulated
in Table B.5. A more viewer-friendly colored chart is also provided as Figure 4.2 on the
following page with the numbers showing the linewidths under the according fabricating
conditions.
4.3 Discussion
The colored chart provides a clear pattern within the data. The linewidth of DLW
depends on laser power, writing speed, and inhibitor concentration in the photoresist. In
all four photoresists, the color of the blocks turns “colder” as we go down the rows for
higher laser power, and turns “warmer” as we go right of the columns for faster writing
speed. As we increase the inhibitor concentration, the color of the chart turns “warmer”.
CHAPTER 4. LINEWIDTH CHARACTERIZATION 32
20
30
40
50
100
150
200
250
Speed HΜmsL
Power HnWL
1.8
0.9
0.7
0.6
2.5
2.0
1.4
1.1
3.7
3.2
2.9
1.6
4.5
3.2
2.6
1.7
(a) Photoresist #1
20
30
40
50
100
150
200
250
Speed HΜmsL
Power HnWL
1.3
0.9
0.6
0.0
1.4
1.1
0.7
0.5
2.0
1.3
1.0
0.8
2.1
1.4
1.0
0.7
(b) Photoresist #2
20
30
40
50
100
150
200
250
Speed HΜmsL
Power HnWL
1.0
0.0
0.0
0.0
1.1
1.0
0.9
0.8
1.6
0.9
0.6
0.4
2.1
1.4
1.3
1.1
(c) Photoresist #3
20
30
40
50
100
150
200
250
Speed HΜmsL
Power HnWL
0.0
0.0
0.0
0.0
1.1
0.9
0.6
0.6
1.1
0.8
0.7
0.6
2.0
1.5
1.1
0.9
(d) Photoresist #4
0
1
2
3
4
5
Μm
(e) Linewidth Calibration
Figure 4.2: Linewidths under Different Conditions
CHAPTER 4. LINEWIDTH CHARACTERIZATION 33
The colored chart provides a table of reference for future fabrications of lines with desired
linewidth.
4.3.1 Laser power/writing speed dependence
Generally in the colored chart, as we go down each row, the color of the blocks turns
“colder”, and the linewidth increases. It makes sense that thicker lines are fabricated with
higher laser power.
All the rows in the colored chart have linewidth value decreasing from left to right.
Thus the linewidth decreases as we increase the writing speed, because the exposure dose
is reduced in a unit area with a faster speed.
4.3.2 Inhibitor concentration dependence
The general picture shows that the color of the charts turns “warmer” as the concen-
tration of inhibitor is increased. Here we compare the linewidths under one power-speed
combination for four different photoresists to illustrate the inhibitor’s affects on DLW
linewidth. Figure 4.3 shows the average linewidths of straight lines fabricated under 150
nW laser power and 20 µm/s writing speed with four different photoresists. As we may
tell from the figure, the average linewidth measured from the SEM image decreases with
an increasing inhibitor concentration. However, we are more interested in the linewidths
before shrinkage, which can be calculated with the shrinkage coefficient determined in
Section 3.2. The linewidth before shrinkage for each photoresist is tabulated in Table 4.2,
CHAPTER 4. LINEWIDTH CHARACTERIZATION 34
0
1
2
3
4
0%
2%
3%
4%
Inhibitor Conc.
Μm
Line Width
2.55
1.39
1.15
1.06
Figure 4.3: Linewidths under 150 nW, 20µm/s with different photoresists
Photoresist Measured linewidth % Shrinkage Calculated linewidth
#1 2.55 µm 20% 3.19 µm
#2 1.39 µm 22% 1.78 µm
#3 1.15 µm 25% 1.53 µm
#4 1.06 µm 28% 1.47 µm
Table 4.2: Calculated linewidth from shrinkage coefficients
0
1
2
3
4
0%
2%
3%
4%
Inhibitor Conc.
Μm
Line Width
3.18
1.79
1.53
1.47
Figure 4.4: Calculated linewidths with different photoresists
and plotted in Figure 4.4. We arrive at the same conclusion that the average linewidth
decreases with an increasing inhibitor concentration.
Part II
Mathematical Modeling
Chapter 5
Introduction II
One of the most important concerns in the area of direct laser writing (DLW) lithog-
raphy is the spatial resolution. However, the spatial resolution of DLW lighography is
affected by the memory effect. The diffusive behavior of the photoresist prohibits struc-
tures from being fabricated close to each other. The heat equation, which is classified as
a parabolic partial differential equation, describes the heat flow or diffusion processes. A
simple example of the heat equation in one-dimension is given by
∂
∂t
u(x, t) = α ·
∂
2
∂x
2
u(x, t), (5.1)
where u(x, t) is a function of space x and time t that describes the concentration of a
specific molecule in a heterogeneous mixture, where α > 0 is the diffusivity coefficient.
Previous research showed the utility of heat equations in this type of modeling [PY05].
We assume that the diffusion of the particles, including the monomer, photoinitiator, and
36
CHAPTER 5. INTRODUCTION II 37
free radicals, behave according to an analog of the heat equation. Therefore, it is natural
to start with the heat equation and develop a mathematical model for the polymerization
of the photoresist.
However, a PDE does not always have analytic solutions [Rub74]. In order to obtain
numerical approximations of the solutions to PDEs, the method of finite differences is used
to discretize the continuous partial differential equations into difference equations. The
difference equations can then be iteratively solved with appropriate initial and boundary
conditions with the help of mathematical programming languages.
Additionally, we use Markov chains to model the growth of the polymer. Similar
approach had been applied in the modeling of actin polymerization [Sza10]. A Markov
chain is a deterministic probabilistic model with a finite number of states {s
i
}
n
i=1
and
transition probabilities {p
ij
| 1 ≤ i, j ≤ n} between states s
i
and s
j
. The Markov
process assumes that the conditional probability distribution of future states of the process
depends only upon the present state, not on the sequence of events that preceded it. The
Markov model gives a probabilistic distribution of states at any given number of time
steps [Rev84].
5.1 The Gaussian Beam
Before we start modeling the polymerization reaction, we study the profile of the laser
setup. In Section 2.1, we proposed that the intensity of the collimated laser beam follows
CHAPTER 5. INTRODUCTION II 38
the pattern of an Airy disk, which can be approximated by a Gaussian function. The
beam can therefore be approximated as a Gaussian beam (Section 2.1 and Figure 2.2).
We shall present some useful properties of the Gaussian beam [ST91].
The optical intensity of a focused Gaussian beam I(r) is a function of the axial and
radial distances, z and ρ =
p
x
2
+ y
2
, respectively. The expression is given by
I(ρ, z) = I
0
W
0
W (z)
2
Exp
−
2ρ
2
W (z)
2
, (5.2)
where I
0
is the intensity at the center of the beam, W
0
is the beam waist, and W (z) is the
beam radius at z, given by
W (z) = W
0
s
1 +
z
z
0
2
, (5.3)
where z
0
is called the Raleigh range, and is the axial distance at which the intensity is
half of I
0
along the beam axis.
Two additional properties of the Gaussian beam used in developing the model are the
beam divergence and the depth of focus. Far from the beam center, when z z
0
, the
beam radius increases approximately linearly. This fact can be verified by examing the
beam radius function W (z). By algebraic manipulation, we may get the relation
W (z)
2
W
2
0
−
z
2
z
2
0
= 1, (5.4)
which implies that W (z) is the positive branch of a pair of hyperbolae. The plot of W (z),
therefore, has a pair of linear asymptotes (Figure 5.1), which defines a half-angle. About
86% of the beam power is confined with a cone defined by the half-angle θ
0
[ST91]. The
CHAPTER 5. INTRODUCTION II 39
0
z
0
-z
0
z
W
0
2 W
0
WHzL
Θ
0
Figure 5.1: Beam Radius as a Function of z
angular divergence of the beam is therefore defined by the angle
θ
0
=
2
π
·
λ
2W
0
. (5.5)
The beam divergence is inversely proportional to the beam-waist diameter 2W
0
. Thus if
we want to squeeze the beam waist to produce a smaller DLW voxel, a larger θ
0
should
be used by using a lens with high numerical aperture (NA).
The Gaussian beam achieves its best focus at the plane z = 0. In either direction, the
beam gradually grows “out of focus.” The axial distance within which the beam radius
lies within a factor of
√
2 of its minimum value is known as the depth of focus or confocal
parameter, and is twice the Rayleigh range,
2z
0
=
2πW
2
0
λ
. (5.6)
In our 40× oil setup (See Section 2.5), we know the numerical aperture (NA) of the
objective lens and the refractive index n of the index matched immersion oil. Thus we
compute the angular divergence of the Gaussian beam
θ
0
= sin
−1
NA
n
. (5.7)
CHAPTER 5. INTRODUCTION II 40
With the help of the angular divergence formula (5.5), we may figure out the beam waist
W
0
=
λ
πθ
0
=
λ
π sin
−1
(NA/n)
(5.8)
Now we substitute the numerical values λ = 0.405 µm, NA = 1.3, and n = 1.5 in, and
obtain the beam waist W
0
= 0.123 µm. Thus the intensity function (5.2) at z = 0 is
I(r, θ) = I
0
· Exp
−
2r
2
W
2
0
, where W
0
= 0.123 µm. (5.9)
We integrate the intensity function over the whole plane in polar coordinates to get the
power output of the Gaussian beam:
P =
ZZ
R
2
I(r, θ)dA =
Z
2π
θ=0
Z
∞
r=0
I
0
· Exp
−
2r
2
W
2
0
· rdrdθ =
πI
0
W
2
0
2
. (5.10)
Therefore, once we measure the power output P of the approximated Gaussian beam after
the pinhole, we may use the equation
I
0
=
2P
πW
2
0
(5.11)
to compute the principal intensity. Thus we have the complete Gaussian beam profile,
with the radial distance r and axial distance z from the center of exposure as the only
two dependent variables.
5.2 Parabolic Equations
To describe diffusive behaviors of particles, we need parabolic-type partial differential
equations. First we shall have some definitions [Far82].
CHAPTER 5. INTRODUCTION II 41
Definition 1. A partial differential equation, abbreviated as PDE, is an equation
that contains partial derivatives.
Let u(x, y) be a function of x and y. Listed below are some examples of partial
differential equations.
∂u
∂x
= 0,
∂
2
u
∂x
2
+
∂
2
u
∂y
2
= 0, u
2
+
∂u
∂x
= f(x, y).
For notational simplicity, we would denote the partial derivatives using subscripts as
follows
u
x
=
∂u
∂x
, u
y
=
∂u
∂y
, u
xx
=
∂
2
u
∂x
2
, u
yy
=
∂
2
u
∂y
2
.
Definition 2. Let u(x, y) be a function of two variables x and y. A second-order linear
PDE in two variables is an equation in the form
Au
xx
+ Bu
xy
+ Cu
yy
+ Du
x
+ Eu
y
+ Fu = G,
where A, B, C, D, E, F, G can be constants or given functions of x and y.
Definition 3. Parabolic equations are second-order linear PDEs which satisfy the
property B
2
−4AC = 0. Parabolic PDEs are often used to describe heat flow and diffusion
processes.
Definition 4. The one-dimensional heat equation is a PDE in the form
u
t
= α ·u
xx
,
where u is a function of space x and time t, and α > 0 is the diffusivity coefficient.
CHAPTER 5. INTRODUCTION II 42
The heat equation is parabolic because 0
2
−4α ·0 = 0, and it is capable of describing
the diffusion of particles in a system. However, the heat equation only describes linear
diffusive behaviors under conservative conditions. The introduction of source and sink
terms enables us to incorporate the non-conservative behaviors of concentrations. Notice
that the left-hand side of the heat equation is the partial derivative of the concentration
function with respect to time, in other words, the rate of change in concentration. A
positively-valued term added to the right-hand side leads to the increase in u
t
, and we
name such term a source. On the other hand, a negatively-valued term is named a
sink. These terms describes the non-conservative increase or decrease in concentrations
of species in the system.
In addition, to solve differential equations, we need initial conditions and boundary
conditions. The initial condition provides information about the concentration in the
system at an initial time. The boundary conditions are given to compensate for the fact
that the partial derivatives are not well-defined at the boundaries. During the diffusion
process, there should be no net flow of particles through the boundary, i.e., the spatial
derivatives at the boundaries should be zero. This type of boundary conditions is called
Neumann boundary conditions.
With an appropriate system of parabolic PDEs, initial, and boundary conditions, we
shall proceed to numerically solve the system with the method of finite differences, since
it would be too hard, if not impossible, to solve analytically.
CHAPTER 5. INTRODUCTION II 43
5.3 Method of Finite Differences
The differential equations with sink and source terms described in the previous section
are usually nonlinear, and thus too complicated to be solved analytically. We need to de-
vise our toolbox with numerical methods to solve them. Of the many different approaches
to solving PDEs numerically, we use the method of finite differences to numerically approx-
imate our physical reality [Tho95]. The method of finite differences converts differential
equations to difference equations, whose spatial and time domains are discrete. There
are yet other numerical methods that are more efficient and accurate, but the method
of finite differences is easier to implement in a programming language. We use Wolfram
Mathematica
R
to numerically solve the PDEs which simulate the physical problem.
To start, we consider the initial-boundary-value problem of heat flow. Let V (x, t) be
a function of temperature in terms of space x and time t on the domain x ∈ [0, 1]. Let
f(x) be a function defined on x ∈ [0, 1], and let a(t) and b(t) be functions defined on
t ∈ [0, ∞). Suppose that V (x, t) satisfies
V
t
= α ·V
xx
, x ∈ (0, 1), t > 0 (5.12)
V (x, 0) = f(x), x ∈ [0, 1] (5.13)
V (0, t) = a(t), V (1, t) = b(t), t ≥ 0 (5.14)
where the initial condition and the boundary conditions agree, i.e., f(0) = a(0) and
f(1) = b(0). Our first step is to convert the continuous system into a discrete expression.
We discretize the spatial domain into M + 1 equally spaced points x
0
, x
1
, . . . , x
M−1
, x
M
,
CHAPTER 5. INTRODUCTION II 44
where x
0
= 0, x
M
= 1, and x
k
= k∆x for all integer-valued k that satisfies 0 ≤ k ≤ M
and ∆x = 1/M. Likewise, we discretize the time domain with spacing ∆t, and we have
points t
0
, t
1
, . . . on the time domain, where t
n
= n∆t for all integer-valued n ≥ 0. Hence,
we obtain a grid on the space-time domain as shown in Figure 5.2.
x
t
x
0
x
1
x
2
x
M-1
x
M
D x
D t
2 D t
3 D t
4 D t
0
Figure 5.2: Difference Grid on the Space-Time Domain
The lattice points in the figure approximate the system (5.12)–(5.14) as ∆x → 0 and
∆t → 0, and we shall define v(k, n) to be a function defined at the point (k∆x, n∆t), in
which case v approximates V . By definition of a derivative, we know that
V
t
(x, t) = lim
∆t→0
V (x, t + ∆t) − V (x, t)
∆t
. (5.15)
Thus it is reasonable to define the derivative of the discrete function v(k, n) by
v
t
(k, n) =
v(k, n + 1) − v(k, n)
∆t
. (5.16)
Likewise, the second spatial derivative of v(k, n) may be defined by the centered difference
CHAPTER 5. INTRODUCTION II 45
equation
v
xx
(k, n) =
v(k + 1, n) − 2v(k, n) + v(k −1, n)
∆x
2
. (5.17)
Thus the approximated differential equation (5.12) is written as
v(k, n + 1) − v(k, n)
∆t
= α ·
v(k + 1, n) − 2v(k, n) + v(k −1, n)
∆x
2
, (5.18)
or, equivalently,
v(k, n + 1) = v(k, n) + α ·
∆t
∆x
2
·
v(k + 1, n) − 2v(k, n) + v(k −1, n)
, (5.19)
and the initial condition and the boundary conditions are approximated by
v(k, 0) = f(k∆x) (5.20)
v(0, n + 1) = a((n + 1)∆t) (5.21)
v(M, n + 1) = b((n + 1)∆t) (5.22)
The equations (5.19)–(5.22) provide an approximation for the system (5.12)–(5.14). Be-
cause we already know the initial values of v at all lattice points and the boundary values
of v at all time from equations (5.20)–(5.22), we may use equation (5.19) to iteratively
compute the value of v at each lattice point, thus giving a discrete numerical approxima-
tion for the continuous function V (x, t).
5.4 Markov Chains
A probabilistic model for the acrylate polymerization can be presented as a directional
random walk [Sza10]. A random walk is a random process consisting of a sequence of
CHAPTER 5. INTRODUCTION II 46
discrete steps. Consider a particle on the number line with initial location x = 0. In
a directional random work, there are two and only two possible options for the particle
in a time unit: there is a probability p that the particle jumps one unit to the right,
and a probability q that it stays at its current location. Additionally, the probabilities
must sum to one, i.e., p + q = 1. The states (position of the particle) and the transition
probabilities between states together form a Markov chain as shown by the state diagram
in Figure 5.3. The Markov chain satisfies the Markov assumption that the probability of
transition is only dependent on the current state and independent from all previous ones,
thus can be modeled as a Markov process.
0
p
0
1 - p
0
1
p
1
1 - p
1
2
p
2
1 - p
2
3
p
3
1 - p
3
Figure 5.3: A State Diagram of the Markov Chain for Directional Random Walk
Define P (k, t) for all k ∈ N and t ∈ N as the probability of the particle being at
position k at time t, and let p
k
be the transition probability for a transition from k to
k + 1 for all k ∈ N. Given an initial probability density function P (k, 0) of where the
particle might be, the probability density P (k, i + 1) at the (i + 1)-th time step can be
given by
P (k, i + 1) = P (k, i) · (1 −p
k
) + P (k − 1, i) ·p
k−1
. (5.23)
The first term on the right hand side stands for the probability that the particle was at
CHAPTER 5. INTRODUCTION II 47
position k at the i-th time step, and did not move in the subsequent time iteration; the
second term stands for the probability that the particle was at position k − 1 at the i-th
time step, and moved one step forward in the subsequent time iteration. Mathematical
induction allows us to iteratively approximate the probability density P(k, t) at the t-th
time step for all k ∈ N. The expected value E[K, t] of the location of the particle at time
t can be computed with
E[K, t] =
∞
X
k=0
k · P (k, t). (5.24)
We may use this expected value as the expected position of the particle at time t.
Chapter 6
Memory Effect Characterization
The ultimate goal of this project is to develop a mathematical model that describes
the acrylate polymerization reaction as well as the memory effect (Section 1.4). Here we
present an experimental method to quantify the memory effect. We studied the simplest
case – two dots drawn consecutively with a hiatus time in between. Because both the
diffusive behavior and the decay of radicals depend heavily on time, we need to accurately
measure all the time periods between and during events. In addition to our usual setup, we
used an oscilloscope to measure the exposure time and hiatus time. The percent growth
in dot size is measured and tabulated for different spacing and hiatus time conditions,
and is used as a measure of the “extent” of memory effect.
48
CHAPTER 6. MEMORY EFFECT CHARACTERIZATION 49
6.1 Oscilloscope Setup
A digital oscilloscope (Rigor DS1204B) was connected to the shutter control to monitor
the electronic signals. The voltage output of the shutter trigger was connected to Channel
1 of the oscilloscope. The time scale was set to be 500.0 ms per grid, and the voltage
scale was set to be 20.0 V per grid (Figure 6.1a). For monitoring a double-dot exposure,
the oscilloscope run control was set to “Single”, which is triggered by a change in the
input voltage. A measurement of 0.50-second exposure time and 1.0-second hiatus time
is shown in Figure 6.1b.
(a) Oscilloscope Setup (b) A measurement by oscilloscope
Figure 6.1: Digital oscilloscope used in the experiment
6.2 Methods
A sample for DLW is prepared as described in Section 2.5. Then pairs of dots with
0.50-second exposure time were fabricated with spacing and hiatus time listed in Table 6.1.
CHAPTER 6. MEMORY EFFECT CHARACTERIZATION 50
Spacing (µm) Hiatus time (s) Base speed range (µm/s)
10.0 2.0 5.7 – 5.9
10.0 1.0 11.5 – 11.7
10.0 0.50 24.3 – 24.6
8.0 2.0 4.8 – 5.0
8.0 1.0 10.2 – 10.5
8.0 0.50 21.8 – 22.1
5.0 2.0 3.1 – 3.3
5.0 1.0 6.3 – 6.6
5.0 0.50 13.8 – 14.1
Table 6.1: Base Speed Range for Specific Hiatus Time
Depending on the amount of index matched immersion oil applied between the sample
and the lens, the actual moving speed of the automated stage differs from the base speed
set by the DLW program. The base speed of the program was adjusted according to
Table 6.1 so that the hiatus time can be as close to the desired value as possible. At
the same time, a camera (Thorlabs UC480) mounted on the microscope was set to a
resolution of 480 × 240 pixels with a frame rate of 101.61 frames per second for further
high-speed analysis in the model. The laser power was set to 200 nW. When the laser,
program, oscilloscope, and camera were all prepared accordingly, a double-dot fabrication
was operated with the camera taking a high-speed video. After all the double-dots were
CHAPTER 6. MEMORY EFFECT CHARACTERIZATION 51
completed, the sample was washed in ethanol and dried with gentle breeze of air. It was
then sputter-coated with gold and sent for SEM imaging, where we obtain data for the
dot sizes in each double-dot fabrication.
6.3 Results
The size of each dot under each fabrication are listed in Table 6.2. Notice that some
of the hiatus times or exposure times were not controlled exactly the same as the desired
value because the machine sometimes had unpredictable electronic delays. In addition,
the last two sets (#8, #9) of measurements are not provided because the two dots merged
with each other due to the lack of space between them. Figure 6.2 shows the outcome
of measurement #9. The SEM images of all nine sets of double-dot measurement are
provided in Appendix C.
Figure 6.2: SEM Image of Double-Dot Experiment #9
CHAPTER 6. MEMORY EFFECT CHARACTERIZATION 52
#
Spacing
(µm)
1
st
exposure
(s)
2
nd
exposure
(s)
Hiatus time
(s)
1
st
Dot
(µm)
2
nd
Dot
(µm) % Increase
1 10 0.50 0.50 2.0 1.81 2.09 15%
2 10 0.50 0.50 1.0 2.07 2.56 24%
3 10 0.50 0.50 0.54 1.99 2.63 32%
4 8 0.50 0.50 2.0 1.80 2.14 19%
5 8 0.50 0.50 1.0 2.09 2.70 29%
6 8 0.50 0.56 0.50 2.30 3.37 47%
7 5 0.50 0.54 2.0 2.70 3.28 21%
8 5 0.50 0.56 1.0 — — —
9 5 0.50 0.50 0.50 — — —
Table 6.2: Dot Size Measurements for Memory Effect
CHAPTER 6. MEMORY EFFECT CHARACTERIZATION 53
6.4 Discussion
From Table 6.2, two points can be concluded:
• if the spacing between two dots is controlled, the shorter the hiatus time, the larger
the percent increase in size;
• if the hiatus time between two dots is controlled, the shorter the distance between
them, the larger the percent increase in size.
This result is within our expectation, because the memory effect is conjectured to be
a result of excessive amount of radicals and oligomer molecules due to non-local exposure
and diffusive behaviors. The diffusion needs time to proceed, thus in a given amount of
time, the further away the second exposure is, the less influence from the first exposure
imposes on the second. The radicals also decay when two of them quench each other
(Section 1.3). The decay process also has a rate, so that the longer we wait after we
fabricate the first dot, the less influence from the first exposure imposes on the second.
Chapter 7
Modeling the Memory Effect
We first modeled the memory effect with the heat equation and approximated the
changes in the concentration of photoinitiator and radicals (including initial radicals,
acrylate monomer radicals, and acrylate oligomer radicals). The PDEs were numerically
solved with the method of finite differences, and the solutions were computed in Mathe-
matica as functions of space and time. We then associated the transition probabilities of
a Markov chain to the radical concentration, and utilized the Markov chain to generate
a probabilistic model that describes the growth of polymer. When the model was com-
pleted, the parameters were adjusted so that the results from the model better fit the
experimental data obtained in Chapter 6.
In order to prevent confusion, a table of variables and coefficients, and a table of
functions and lists, are provided in Tables 7.1 and 7.2.
54
CHAPTER 7. MODELING THE MEMORY EFFECT 55
Variables
in Model
Variables
in Code Definitions and Comments
— ExposureTime1 Variable for the first exposure time in s
— ExposureTime2 Variable for the second exposure time in s
— HiatusTime Variable for the hiatus time in s
— Spacing Variable for the spacing between two dots in µm
λ Lambda Wavelength of laser in µm, specifically, λ = 0.405
P P Power output of the laser in nW, specifically, P = 200
θ Theta Angular divergence of the Gaussian beam
z z Axial distance from the center of the beam in µm
z
0
z0 Raleigh range in µm
W
0
W0 Beam waist in µm
I
0
I0 Principal intensity of the bean in W/m
2
α alpha Diffusivity of photoinitiator
β beta Diffusivity of radicals
m m Proportion of photoinitiator yield to beam intensity
d d Dying coefficient of radicals
M M Total number of spatial grids, specifically, M = 400
∆x dx Length of each spatial step in µm
∆t dt Length of each time step in s, specifically, t = 0.002
k k Index of spatial steps (0 ≤ k ≤ M)
n n Index of time steps (n ≥ 0)
Table 7.1: Table of Variables and Coefficients
CHAPTER 7. MODELING THE MEMORY EFFECT 56
Functions
in Model
Functions
in Code Definitions and Comments
W (z) W[z] Returns the beam radius in µm
I(x) Intensity[x] Returns the beam intensity in W/m
2
R(x) Reflection[x] Returns the reflection intensity in W/m
2
— Exposure[x,pos] Returns the exposure dose given the exposure at pos
p densityP[u] Returns the probability of polymerization given the
concentration u of radicals
c c List storing the photoinitiator concentration
u u List storing the radical concentration
P
l
pLeft List storing the spatial probability distribution of the
left boundary of the polymer
P
r
pRight List storing the spatial probability distribution of the
left boundary of the polymer
— Poly List storing the polymerized spatial grids
Table 7.2: Table of Functions and Lists
7.1 First Approximation: Radical Initiation
The first chemical change we approximate is the consumption of the photoinitiator.
The concentration of the photoinitiator is denoted c(x, t) as a function of space x and
time t. We choose a spatial domain wide enough to enclose all changes necessary for con-
CHAPTER 7. MODELING THE MEMORY EFFECT 57
sideration. In this case, we define our spatial domain with a length of 20 µm, and the first
exposure is set at x = 5 µm. As the photoresist is exposed to the laser beam, the photoini-
tiator gets excited and splits into two radicals, thus reducing its own concentration. We
assume that the amount of excited photoinitiator molecules is proportional to the inten-
sity of the exposure and the concentration of photoinitiator. Because the concentrations
of species are hard, if not impossible, to determine experimentally, we assume arbitrary
units for concentrations. Let α be the diffusivity of photoinitiator in the photoresist, I(x)
be the intensity function, R(x) be the function that describes the intensity of reflections
within the photoresist, and m be the conversion coefficient of photoinitiation with respect
to beam intensity. Then the PDE that describes the concentration of photoinitiator is
c
t
= α ·c
xx
− m ·
I(x) + R(x)
· c, (7.1)
with initial condition
c(x, 0) = 1, (7.2)
and boundary conditions
c
x
(0, t) = c
x
(20, t) = 0. (7.3)
It turns out, in the end, that the reflection term R(x) plays an important role in this
model of memory effect. Without this term, we were not able to obtain data for the
percent increase that was comparable to the experimental results. A plot of the reflection
term, R(x), together with the beam intensity, I(x), is given in Figure 7.1, where the blue
curve shows the beam intensity and the purple curve shows the intensity due to reflection.
CHAPTER 7. MODELING THE MEMORY EFFECT 58
-10
-5
0
5
10
0.00
0.02
0.04
0.06
0.08
0.10
0.12
Figure 7.1: Beam Intensity I(x) and Reflection Intensity R(x)
We convert this PDE into a difference equation with the method introduced in Sec-
tion 5.3. Equation (7.1) can be rewritten into
c(k, n + 1) −c(k, n)
∆t
= α ·
c(k + 1, n) − 2c(k, n) + c(k −1, n)
∆x
2
−m ·
I(k∆x) + R(k∆x)
· c(k, n).
(7.4)
Solving for the future time step,
c(k, n + 1) = c(k, n) + α ·
∆t
∆x
2
·
c(k + 1, n) − 2c(k, n) + c(k −1, n)
−∆t ·m ·
I(k∆x) + R(k∆x)
· c(k, n).
(7.5)
The initial condition (7.2) can be converted into
c(k, 0) = 1. (7.6)
Notice that we use Neumann boundary conditions for this approximation because we want
to assume zero net flow of concentrations at the boundaries. This is different from the
example given in Section 5.3, where we considered Dirichlet boundary conditions. There
are several ways to treat the Neumann boundary conditions [Tho95]. For simplicity, we
CHAPTER 7. MODELING THE MEMORY EFFECT 59
would use a one-sided difference approximation of c
x
(0, t) to get
c(1, n + 1) − c(0, n + 1)
∆x
=
c(M, n + 1) − c(M −1, n + 1)
∆x
= 0,
or, equivalently,
c(0, n + 1) = c(1, n + 1) and c(M, n + 1) = c(M − 1, n + 1). (7.7)
Thus, equations (7.5)–(7.7) approximate the differential equation described by equations
(7.1)–(7.3).
7.2 Second Approximation: Acrylate Initiation
With a method similar to the last section, we can also approximate the concentration
of radicals. This concentration is denoted u(x, t) as a function of space x and time t on the
same spatial domain x ∈ [0, 20]. The mechanism mentioned in Section 1.3 suggests that
one molecule of photoinitiator gets excited and splits into two radicals. Conservation of
Mass implies that the increase in radicals equals twice the consumption in photoinitiators.
At the same time, radicals are not stable and tend to quench each other when they
are placed in the same vessel. Hence, a sink term describing the decay of radicals is
also included. Radical quenching always involve two radicals, which suggests that the
quenching reaction is second-order. Thus the rate of decay, given by d, is proportional to
the second power of radical concentration. Let β be the diffusivity of radicals. Then the
CHAPTER 7. MODELING THE MEMORY EFFECT 60
PDE describing the radical concentration is
u
t
= β ·u
xx
+ 2m ·
I(x) + R(x)
· c − d · u
2
, (7.8)
with initial condition, while no radical should be present initially,
u(x, 0) = 0, (7.9)
and boundary conditions
u
x
(0, t) = u
x
(20, t) = 0. (7.10)
Notice that this PDE involves the concentration of photoinitiator c(x, t). Thus it forms a
coupled system of PDEs with equations (7.1)–(7.3). Its corresponding difference equation
is
u(k, n + 1) −u(k, n)
∆t
= β ·
u(k + 1, n) − 2u(k, n) + u(k −1, n)
∆x
2
+2m ·
I(k∆x) + R(k∆x)
· c(k, n) − d · u
2
(k, n),
(7.11)
or, equivalently,
u(k, n + 1) = u(k, n) + β ·
∆t
∆x
2
·
u(k + 1, n) − 2u(k, n) + u(k −1, n)
+∆t ·2m ·
I(k∆x) + R(k∆x)
· c(k, n) − ∆t · d · u
2
(k, n),
(7.12)
with initial condition
u(k, 0) = 0, (7.13)
and boundary conditions
u(0, n + 1) = u(1, n + 1) and u(M, n + 1) = u(M − 1, n + 1). (7.14)
CHAPTER 7. MODELING THE MEMORY EFFECT 61
Therefore, equations (7.5)–(7.7) and (7.12)–(7.14) form a coupled system of difference
equations that describes the concentration of photoinitiator and radicals in the photore-
sist. Our next step is to model the process of polymerization using Markov chains.
7.3 Probability of Polymerization
We use the concept of Markov chains introduced in Section 5.4 to model the poly-
merization of the photoresist, in which case we consider the particle to be the boundaries
of the polymer. It simplifies the two-dimensional polymerization by modeling along the
one-dimensional line connecting two exposure centers. A jump to the right to be poly-
merization over one spatial unit, and no jump as no growth in the polymer. Then the
states of this Markov chain are the possible positions for the left and right boundaries
of the polymer, and the transition probability is the probability of polymerization given
the radical concentration at the position of the boundary. We first define the transition
probability p(u), i.e., probability of polymerization, as a function of radical concentration
u.
First of all, we expect there to be a threshold on the radical concentration that governs
the polymerization reaction: higher than the threshold concentration, polymerization
is likely to occur, and vice versa. Then we consider the extremes: when the radical
concentration is close to zero, the probability of polymerization is expected to be close
to 0; when the radical concentration is very high, the probability of polymerization is
CHAPTER 7. MODELING THE MEMORY EFFECT 62
expected to be close to 1. Finally, we expect a smooth transition near the threshold
concentration. In conclusion, a probability of polymerization p(u) as a function of radical
concentration u is expected to have a plot resembling Figure 7.2a. Note that the plot of the
probability function resembles the phase paths of the differential equation y
0
= ky(1 −y),
where k is a positive constant.
0.0
0.5
1.0
1.5
2.0
0.0
0.2
0.4
0.6
0.8
1.0
Radical Concentration
Probability
(a) Proposied Probability of Polymerization
0
2
4
6
8
10
12
14
0.0
0.2
0.4
0.6
0.8
1.0
(b) Phase Diagram of y
0
= ky(1 −y)
Figure 7.2: Modeling the Probability Function
The general solutions to the differential equation y
0
= ky(1 − y) is
y(x) =
1
1 + e
c−kx
, (7.15)
where c and k are constants, and k > 0. A derivation of the solutions to this differential
equation is provided in Appendix F.
Now we obtain our probability of polymerization p(u) as a function of radical concen-
tration u:
p(u) =
1
1 + e
c−ku
, (7.16)
CHAPTER 7. MODELING THE MEMORY EFFECT 63
where the constant k controls the “steepness” of the transition between 0 and 1, and the
ratio of c and k controls the threshold concentration for polymerization. We shall proceed
to use this probability of polymerization for Markov modeling.
7.4 Third Approximation: Markov Modeling
The Markov model is based on the difference model described in Sections 7.1 and 7.2.
We define P
l
(k, n) and P
r
(k, n) as the probability of the left and right boundaries, respec-
tively, to be at the k-th spatial unit at the n-th time step, where k is an integer between
0 and M, and n is a non-negative integer. The polymerization is expected to start at the
center of exposure x = 5 µm. It follows that the initial conditions for the probability den-
sity functions of the left and right boundaries, P
l
(k, n) and P
r
(k, n), respectively, should
be defined by
P
l
(k, 0) = P
r
(k, 0) =
1, if k∆x ≤ 5 < (k + 1)∆x,
0, otherwise.
(7.17)
Because most of the cases that we experimentally tested did not yield dot size greater
than 10 µm, it is safe to assume the boundary conditions
P
l
(0, n) = P
l
(M, n) = 0, and P
r
(0, n) = P
r
(M, n) = 0. (7.18)
The discrete probability density functions are updated during each time iteration accord-
ing to equation (5.23). Since we have derived the probability of polymerization with
CHAPTER 7. MODELING THE MEMORY EFFECT 64
respect to radical concentration, we may rewrite the equation into
P
l
(k, n + 1) = P
l
(k, n) ·
1 −p
u(k, n)
+ P
l
(k + 1, n) · p
u(k + 1, n)
, (7.19)
and
P
r
(k, n + 1) = P
r
(k, n) ·
1 −p
u(k, n)
+ P
r
(k − 1, n) · p
u(k − 1, n)
. (7.20)
With equations (7.17)–(7.20), we are able to obtain the discrete probability density func-
tions P
l
(k, n) and P
r
(k, n), which describe the positions of the left and right boundaries
of the polymer at any time step. With the probability density functions, the expected
values of the boundary positions can be computed by equation (5.24). We model the
second exposure in the same way, except that we need to use different initial conditions
in (7.21)–(7.23) according to the spacing between two exposures.
c
second
(k, 0) = c
first
(k, FinalTimeIteration) (7.21)
u
second
(k, 0) = u
first
(k, FinalTimeIteration) (7.22)
P
l
(k, 0) = P
r
(k, 0) =
1, if k∆x ≤ 5 + Spacing < (k + 1)∆x,
0, otherwise.
(7.23)
Significant changes in physical properties take place when the acrylate-based photore-
sist turns into polymer. The index of refraction changes when the photoresist turns from a
liquid to a solid, and thus we are able to see the polymer under the microscope. Moreover,
the diffusivity changes dramatically because it is hard for particles to flow within solid
state matters. The diffusive behaviors are therefore turned off by setting the diffusivity
coefficients α and β to zero between the boundaries of the polymer.
Chapter 8
Mathematica Implementation
In this chapter, we first develop a framework of the implementation in Mathematica
by programming the model for the system of PDEs that describes the concentrations
of photoinitiator and radicals. Then we use the framework to implement the markov
model into the program. The result of the model is compared with the experimentally-
determined data to evaluate the validity of the model.
8.1 Finite Difference Modeling
The program starts by setting up all the parameters and variables for the beam profile
and exposure. The initial conditions of c and u are set with a For loop which generates
two lists initc and initu of length M + 1 consisting of 1’s and 0’s, respectively. These
two lists are appended into c and u, respectively, as the initial time iteration.
65
CHAPTER 8. MATHEMATICA IMPLEMENTATION 66
A new For loop is used for time iterations. At the start of each time iteration, empty
lists cnew and unew are generated for storing the values of concentration during the
current time iteration. Another For loop is then used for spatial iterations, which puts
the concentration values into the lists cnew and unew. An If statement checks whether
the current spatial grid is at the boundary and, if it is, the If statement makes changes
to preserve the boundary conditions. At the end of each time iteration, the completed
lists cnew and unew are appended to lists c and u, respectively.
When all the time iterations are completed, a Manipulate command, together with a
ListLinePlot, is used to show the change in the concentration of species during the time
period which the program simulated.
8.2 Markov Modeling
The implementation of Markov modeling inherits the framework set up in the previous
section. In addition to the initial conditions for c and u, we also need to include the initial
conditions for pLeft and pRight. Moreover, we introduce another list Poly to keep track
of the polymerized spatial units so that we are able to turn off diffusive behaviors in
these units. The content of this list is essentially only 0’s and 1’s, where a 0 means not
polymerized and 1 polymerized for the corresponding spatial grid.
In the For loop for time iteration, two new lists pnewLeft and pnewRight are con-
structed in the same fashion as we construct c and u, and appended to pLeft and pRight,
CHAPTER 8. MATHEMATICA IMPLEMENTATION 67
respectively, at the end of each time iteration. The expected values of left and right
boundaries are calculated after pnewLeft and pnewRight are generated, and an update of
newPoly, which has 1’s between the expected boundaries and 0’s elsewhere, is appended
to Poly. For the second exposure, the initial conditions for c, u, and Poly are the same
as their final states during the simulation for the first exposure, but the initial conditions
for pLeft and pRight have to be reset to the new exposure center because we cannot
have four boundaries for a growing polymer.
The Manipulate command is modified so that the expected boundaries of the polymer
dots are visible in the simulation. In addition, the number of polymerized spatial units
are printed as the output at the end of the simulation for each exposure. The percent
increase is also calculated at the very end of the program.
Pseudocodes for the complete implementation is provided in the next section. Addi-
tionally, a completed version of the program in Wolfram Mathematica
r
8 is provided in
Appendix D. The code is fragmented into functional modules and arranged as sections of
the appendix. The indices of lists in the code are off by 1 because we started our indices
at 0 but Mathematica takes 1 as the index of the first item of a list structure.
8.3 Pseudocode for the Implementation
The pseudocode for modeling the polymerization under a double exposure is included
in the next page to demonstrate the flow of the program.
CHAPTER 8. MATHEMATICA IMPLEMENTATION 68
Parameters
Set up the adjustable parameters
c ← {initc}, u ← {initu},
pLeft ← {initp}, pRight ← {initp}, P oly ← {initP oly},
Set up the initial conditions
for n ← 1, n∆t ≤ F inalT ime do Time iteration
cnew = unew = pnewLeft = pnewRight = {}
for k ← 0, k ≤ M do Spatial iteration
if P oly(k) = 1 then
Set α and β to be zero in this spatial iteration
end if
if k = 0 or k = M then
Update cnew, unew, pnewLeft, pnewRight according to boundary conditions
i.e. equations (7.7), (7.14), and (7.18)
else
Update cnew, unew, pnewLeft, pnewRight according to difference equations
i.e. equations (7.5), (7.12), (7.19), and (7.20)
end if
end for
Append cnew, unew, pnewLeft, pnewRight to c, u, pLeft, pRight, respectively
Compute the expectation values leftExpect and rightExpect of the boundaries
P olynew ← {1 within leftExpect and rightExpect, 0 otherwise}
Append P olynew to P oly
end for
return Dot1 ← # of 1’s in the final state of P oly Dot 1 size
Set up the initial condition for the second exposure
Final state of the first exposure
for n ← 1, n∆t ≤ F inalT ime do
Exactly the same for loop with the first one
end for
return Dot2 ← # of 1’s in the final state of P oly −Dot1 Dot 2 size
return percentIncrease ← (Dot2 − Dot1)/Dot1
Manipulate command that shows the dynamics of polymerization
CHAPTER 8. MATHEMATICA IMPLEMENTATION 69
8.4 Results and Discussion
The parameters of the model were adjusted to fit the experimental data provided in
Table 6.2. Table 8.1 shows the chosen values of all the adjustable parameters and functions
used in the final version of the model. The first dot of each experiment has 0.5-second
exposure, and we assume the second exposure does not affect the size of the first dot.
Parameters or Functions Values of Choice
z 1 µm
α 0.2 µm
2
/s
β 0.1 µm
2
/s
m 400 s
−1
/I
0
d 0.3 AU
−1
·s
−1
R(x)
I(x/20) + I(x/10)
12
p(u)
1
1 + e
20−25u
Table 8.1: Choices for Parameters and Functions
Then we compute the average size of the first dot to be 2.1 ± 0.3 µm. This size was
measured under SEM, thus after shrinkage. If we account for the shrinkage coefficient of
photoresist #3 (25%, Table 3.2), the average dot size before shrinkage would be 2.8 ±0.4
µm. After painstaking efforts on adjusting the parameters, we achieved 2.76 µm for the
size of the first dot computed by the model. The dot sizes simulated by the model are
provided in Table 8.2. The simulation of experiment #2 is provided in Figure 8.1 by one
CHAPTER 8. MATHEMATICA IMPLEMENTATION 70
#
Spacing
(µm)
1
st
exposure
(s)
2
nd
exposure
(s)
Hiatus time
(s)
1
st
Dot
(# grids)
2
nd
Dot
(# grids) % Increase
1 10 0.50 0.50 2.0 69 76 10%
2 10 0.50 0.50 1.0 69 87 26%
3 10 0.50 0.50 0.54 69 98 42%
4 8 0.50 0.50 2.0 69 77 12%
5 8 0.50 0.50 1.0 69 94 36%
6 8 0.50 0.56 0.50 69 118 71%
7 5 0.50 0.54 2.0 69 79 14%
8 5 0.50 0.56 1.0 69 123 78%
9 5 0.50 0.50 0.50 69 — —
Table 8.2: Percent Increase Results Simulated by the Model
CHAPTER 8. MATHEMATICA IMPLEMENTATION 71
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.5s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.5s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.5s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
3.s
Figure 8.1: Simulation Result of Experiment #2
CHAPTER 8. MATHEMATICA IMPLEMENTATION 72
screenshot every 0.5 second. The blue and purple lines in the diagrams indicate the con-
centrations (arbitrary unit) of radicals and photoinitiator, respectively. The black boxes
indicate the size of the growing polymer. A more detailed simulation of this experiment
is provided in Appendix E with one screeshot every 0.1 second.
The percent increase values simulated by the model and the experimentally-determined
values are tabulated in Table 8.3. Both sets of data showed the same pattern. The
memory effect is stronger at places close to the exposed areas, and is reduced far away
from fabricated structures. The memory effect is also stronger immediately after one
exposure, and if we wait a while after an exposure, memory effect can be reduced.
# 1 2 3 4 5 6 7 8 9
Experimental 15% 24% 32% 19% 29% 47% 21% — —
Modeling 10% 26% 42% 12% 36% 71% 14% 78% —
Table 8.3: Comparison between Experiments and the Model
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
3.s
Figure 8.2: Final State of Simulation #8
CHAPTER 8. MATHEMATICA IMPLEMENTATION 73
We notice that the experimental data #8 showed two dots merged together due to
their proximity (Figure C.8), yet the model still suggested two distinct dots. But if we
take a closer look at the final state of the model (Figure 8.2), the distance between two
boundaries is only 1 spatial unit, which converts to 50 nm. This distance is so small that
we can consider the dots as merged together.
8.5 Future Development
First of all, we can tune the parameters and adjustable function in the model more
carefully and more finely to make the outputs of the model closer to the experimental
values. We may also test the model for double-dot experiments with different laser power
and/or different exposure time.
We have successfully used a radical scavenger ingredient for the photoresist to reduce
the linewidth as well as the memory effect. We have also developed a model describing
the polymerization reaction with double dots. The natural step to follow would be mod-
eling memory effect in the photoresist without the radical scavenger. By comparing the
differences in these two models, we may be able to suggest the role of radical scavenger
in reducing linewidth and memory effect.
Another possible direction to go would be taking the model into higher dimensions.
Currently the model describes the polymerization reaction along the line connecting two
exposures, which has only one spatial dimension. It is possible that we can develop a two-
CHAPTER 8. MATHEMATICA IMPLEMENTATION 74
dimensional model which tells the polymerization on a plane. Given specific fabrication
conditions, the two-dimensional model may be able to simulate the outcome of a DLW
experiment before actually running the experiment.
Appendix A
Linewidths before/after Alcohol
Development
75
APPENDIX A. LINEWIDTHS BEFORE/AFTER ALCOHOL DEVELOPMENT 76
Linewidth Measurements of Photoresist #1
Before (µm) 46.8 50.2 53.6 58.1 62.7 65.9 69.9 73.8
After (µm) 34.63 37.94 41.33 45.02 49.57 52.77 56.06 59.38
Before (µm) 77.3 78.6 82.2 85.8 89.9 93.0 97.7 101.0
After (µm) 62.28 63.04 66.38 69.64 72.90 76.12 80.59 83.51
Before (µm) 104.9 108.3 114.0 117.7 121.8 124.8 137.7 141.8
After (µm) 86.86 89.50 94.22 97.21 99.15 102.71 123.17 127.42
Before (µm) 145.1 149.4 153.6 154.4 159.7 163.5 166.9 170.5
After (µm) 130.94 134.60 138.23 138.61 142.87 146.39 150.97 153.02
Before (µm) 175.4 179.5 183.9 187.1
After (µm) 156.56 160.88 163.65 167.08
Table A.1: Linewidths before and after developing (photoresist #1)
Linewidth Measurements of Photoresist #2
Before (µm) 60.7 64.5 69.7 72.8 77.1 78.5 81.7 85.5
After (µm) 48.46 52.55 56.54 59.81 63.56 64.77 68.70 72.40
Before (µm) 89.4 94.1 94.4 98.9 102.2 106.6 111.6 111.0
After (µm) 75.90 80.14 80.58 84.81 88.84 92.01 96.02 95.88
Before (µm) 114.9 119.2 122.4 125.9 127.0 131.4 135.5 139.2
After (µm) 99.66 103.56 107.14 109.94 111.36 114.94 118.50 121.97
Before (µm) 142.4 143.9 148.7 151.6 155.1 157.5 158.7 161.9
After (µm) 124.91 126.32 130.53 133.66 137.14 139.84 141.27 144.29
Before (µm) 166.2 169.0 173.0 174.9 177.7 182.0 185.8 189.8
After (µm) 147.83 150.16 153.46 155.28 157.78 161.31 165.49 168.12
Table A.2: Linewidths before and after developing (photoresist #2)
APPENDIX A. LINEWIDTHS BEFORE/AFTER ALCOHOL DEVELOPMENT 77
Linewidth Measurements of Photoresist #3
Before (µm) 57.5 61.2 64.5 68.4 71.8 73.4 78.1 81.0
After (µm) 46.63 50.89 54.02 56.90 59.56 62.33 66.18 69.23
Before (µm) 84.8 87.4 89.7 93.7 97.8 100.8 104.2 106.1
After (µm) 72.42 75.01 77.13 81.13 84.15 87.66 89.95 91.71
Before (µm) 110.1 113.8 117.5 120.8 122.8 125.3 129.4 133.3
After (µm) 95.66 98.88 102.43 103.98 106.72 108.67 113.05 116.40
Before (µm) 136.7 138.7 142.3 146.3 149.9 153.2 154.5 157.7
After (µm) 117.32 119.46 122.38 128.05 129.86 131.53 131.24 135.63
Before (µm) 162.0 165.2 168.1 169.4 173.4 177.1 180.5 183.9
After (µm) 138.68 141.18 143.01 143.45 146.68 149.52 153.05 155.39
Table A.3: Linewidths before and after developing (photoresist #3)
Linewidth Measurements of Photoresist #4
Before (µm) 57.8 61.8 65.4 68.6 71.6 73.3 78.0 81.9
After (µm) 46.09 50.05 53.17 56.11 57.79 59.90 63.80 66.66
Before (µm) 85.4 87.9 90.5 94.4 98.1 101.9 104.4 106.8
After (µm) 70.27 71.70 74.60 78.03 80.40 84.24 85.82 87.00
Before (µm) 110.6 114.7 117.9 120.7 122.6 126.7 130.4 133.4
After (µm) 91.60 94.30 97.48 97.82 99.63 104.44 107.03 109.03
Before (µm) 135.8 138.0 142.9 146.1 149.3 151.6 154.6 158.4
After (µm) 109.32 110.91 115.74 117.61 118.54 119.85 121.75 126.30
Before (µm) 162.4 165.5 167.6 170.1 173.8 177.6
After (µm) 128.32 130.98 128.46 132.62 138.63 140.81
Table A.4: Linewidths before and after developing (photoresist #4)
APPENDIX A. LINEWIDTHS BEFORE/AFTER ALCOHOL DEVELOPMENT 78
y = 0.870413 x
0
50
100
150
200
0
50
100
150
BeforeΜm
AfterΜm
Figure A.1: Linewidth after v.s. before developing (photoresist #1)
APPENDIX A. LINEWIDTHS BEFORE/AFTER ALCOHOL DEVELOPMENT 79
y = 0.876853 x
0
50
100
150
200
0
50
100
150
BeforeΜm
AfterΜm
Figure A.2: Linewidth after v.s. before developing (photoresist #2)
APPENDIX A. LINEWIDTHS BEFORE/AFTER ALCOHOL DEVELOPMENT 80
y = 0.856793 x
0
50
100
150
200
0
50
100
150
BeforeΜm
AfterΜm
Figure A.3: Linewidth after v.s. before developing (photoresist #3)
APPENDIX A. LINEWIDTHS BEFORE/AFTER ALCOHOL DEVELOPMENT 81
y = 0.803336 x
0
50
100
150
200
0
50
100
150
BeforeΜm
AfterΜm
Figure A.4: Linewidth after v.s. before developing (photoresist #4)
Appendix B
Linewidth Data with Different DLW
Conditions
82
APPENDIX B. LINEWIDTH DATA WITH DIFFERENT DLW CONDITIONS 83
Linewidth Measurement Data for Photoresist #1
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
100 nW
#1 1.39 0.78 0.64 0.57
#2 1.66 0.88 0.65 0.55
#3 2.07 0.83 0.72 0.54
#4 1.94 0.87 0.73 0.56
#5 1.92 0.92 0.68 0.59
Average 1.80 0.86 0.68 0.56
(a) 100 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
150 nW
#1 2.27 1.88 1.46 1.16
#2 2.66 1.89 1.31 1.10
#3 2.68 1.98 1.33 1.14
#4 2.59 2.15 1.47 1.09
#5 2.54 2.21 1.56 1.18
Average 2.55 2.02 1.43 1.13
(b) 150 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
200 nW
#1 3.66 3.03 2.72 1.69
#2 3.81 3.10 2.93 1.54
#3 3.83 3.31 2.97 1.60
#4 3.58 3.37 2.78 1.64
#5 3.77 3.41 2.97 1.56
Average 3.73 3.24 2.87 1.61
(c) 200 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
250 nW
#1 4.35 3.19 2.44 1.58
#2 4.46 3.10 2.63 1.67
#3 4.36 3.32 2.47 1.62
#4 4.39 3.25 2.46 1.68
#5 4.84 3.29 2.76 1.75
Average 4.48 3.23 2.55 1.66
(d) 250 nW
Table B.1: Linewidth Measurement Data for Photoresist #1
APPENDIX B. LINEWIDTH DATA WITH DIFFERENT DLW CONDITIONS 84
Linewidth Measurement Data for Photoresist #2
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
100 nW
#1 1.34 0.90 0.58 N/A
#2 1.40 1.04 0.66 N/A
#3 1.34 0.97 0.61 N/A
#4 1.23 0.88 0.50 N/A
#5 1.22 0.87 0.53 N/A
Average 1.31 0.93 0.58 N/A
(a) 100 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
150 nW
#1 1.28 1.09 0.62 0.54
#2 1.38 1.37 0.79 0.59
#3 1.43 1.08 0.72 0.56
#4 1.44 0.89 0.63 0.51
#5 1.44 1.01 0.66 0.54
Average 1.39 1.09 0.68 0.55
(b) 150 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
200 nW
#1 1.89 1.29 0.98 0.81
#2 2.06 1.44 1.00 0.77
#3 2.10 1.35 0.97 0.80
#4 1.97 1.32 0.98 0.80
#5 1.98 1.31 0.94 0.81
Average 2.00 1.34 0.97 0.80
(c) 200 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
250 nW
#1 2.00 1.30 1.04 0.61
#2 1.98 1.52 0.96 0.70
#3 2.26 1.35 0.98 0.66
#4 2.17 1.38 0.94 0.66
#5 2.17 1.37 0.95 0.68
Average 2.12 1.38 0.97 0.66
(d) 250 nW
Table B.2: Linewidth Measurement Data for Photoresist #2
APPENDIX B. LINEWIDTH DATA WITH DIFFERENT DLW CONDITIONS 85
Linewidth Measurement Data for Photoresist #3
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
100 nW
#1 0.91 N/A N/A N/A
#2 1.05 N/A N/A N/A
#3 1.06 N/A N/A N/A
#4 0.83 N/A N/A N/A
#5 0.92 N/A N/A N/A
Average 0.95 N/A N/A N/A
(a) 100 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
150 nW
#1 1.05 0.92 0.83 0.80
#2 1.21 0.95 0.92 0.78
#3 1.24 0.98 0.93 0.83
#4 1.14 0.95 0.80 0.72
#5 1.10 0.98 0.83 0.70
Average 1.15 0.96 0.86 0.77
(b) 150 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
200 nW
#1 1.50 0.88 0.55 0.41
#2 1.51 0.97 0.66 0.44
#3 1.63 1.00 0.60 0.39
#4 1.79 0.91 0.60 0.41
#5 1.67 0.93 0.58 0.44
Average 1.62 0.94 0.60 0.42
(c) 200 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
250 nW
#1 2.04 1.40 1.20 1.13
#2 2.03 1.48 1.19 1.06
#3 2.08 1.37 1.27 0.99
#4 2.04 1.51 1.38 1.07
#5 2.12 1.46 1.31 1.02
Average 2.06 1.44 1.27 1.05
(d) 250 nW
Table B.3: Linewidth Measurement Data for Photoresist #3
APPENDIX B. LINEWIDTH DATA WITH DIFFERENT DLW CONDITIONS 86
Linewidth Measurement Data for Photoresist #4
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
100 nW
#1 N/A N/A N/A N/A
#2 N/A N/A N/A N/A
#3 N/A N/A N/A N/A
#4 N/A N/A N/A N/A
#5 N/A N/A N/A N/A
Average N/A N/A N/A N/A
(a) 100 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
150 nW
#1 0.96 0.89 0.56 0.57
#2 1.04 0.81 0.62 0.58
#3 1.03 0.85 0.66 0.59
#4 1.12 0.94 0.68 0.67
#5 1.14 0.92 0.72 0.69
Average 1.06 0.88 0.65 0.62
(b) 150 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
200 nW
#1 0.97 0.72 0.70 0.59
#2 1.15 0.82 0.59 0.62
#3 1.18 0.84 0.74 0.66
#4 1.16 0.83 0.64 0.51
#5 1.16 0.95 0.71 0.69
Average 1.12 0.83 0.68 0.61
(c) 200 nW
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
250 nW
#1 2.04 1.32 1.03 0.84
#2 2.06 1.42 1.12 0.80
#3 1.86 1.58 1.25 0.91
#4 1.93 1.61 1.20 0.86
#5 1.97 1.37 1.09 0.85
Average 1.97 1.46 1.14 0.85
(d) 250 nW
Table B.4: Linewidth Measurement Data for Photoresist #4
APPENDIX B. LINEWIDTH DATA WITH DIFFERENT DLW CONDITIONS 87
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
Power (nW)
100 1.80 0.86 0.68 0.56
150 2.55 2.02 1.43 1.13
200 3.73 3.24 2.87 1.61
250 4.48 3.23 2.55 1.66
(a) Photoresist #1
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
Power (nW)
100 1.31 0.93 0.58 n/a
150 1.39 1.09 0.68 0.55
200 2.00 1.34 0.97 0.80
250 2.12 1.38 0.97 0.66
(b) Photoresist #2
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
Power (nW)
100 0.95 n/a n/a n/a
150 1.15 0.96 0.86 0.77
200 1.62 0.94 0.60 0.42
250 2.06 1.44 1.27 1.05
(c) Photoresist #3
Linewidth
(µm)
Speed (µm/s)
20 30 40 50
Power (nW)
100 n/a n/a n/a n/a
150 1.06 0.88 0.65 0.62
200 1.12 0.83 0.68 0.61
250 1.97 1.46 1.14 0.85
(d) Photoresist #4
Table B.5: Linewidth Obtained from Different Fabricating Conditions
Appendix C
SEM Images of Double Dots
Figure C.1: SEM Image of Double-Dot Experiment #1
88
APPENDIX C. SEM IMAGES OF DOUBLE DOTS 89
Figure C.2: SEM Image of Double-Dot Experiment #2
Figure C.3: SEM Image of Double-Dot Experiment #3
APPENDIX C. SEM IMAGES OF DOUBLE DOTS 90
Figure C.4: SEM Image of Double-Dot Experiment #4
Figure C.5: SEM Image of Double-Dot Experiment #5
APPENDIX C. SEM IMAGES OF DOUBLE DOTS 91
Figure C.6: SEM Image of Double-Dot Experiment #6
Figure C.7: SEM Image of Double-Dot Experiment #7
APPENDIX C. SEM IMAGES OF DOUBLE DOTS 92
Figure C.8: SEM Image of Double-Dot Experiment #8
Figure C.9: SEM Image of Double-Dot Experiment #9
Appendix D
Coding in Mathematica
D.1 Parameters
(* Experimentally Controlled Variables *)
ExposureTime1 = 0.5; ExposureTime2 = 0.5; HiatusTime = 1; Spacing = 10;
(* Beam Profile Calculation *)
Theta = ArcSin[1.3/1.5]; Lambda = 0.405; W0 =
Lambda/(Pi*Theta); z0 = (Pi*W0^2)/Lambda;(* Beam Profile *)
W[z_] = W0*Sqrt[1 + (z/z0)^2];(* Beam Radius (um) *)
P = 200; (* Power output of laser (nW) *)
I0 = (2 P)/(Pi*W0^2*10^3); (* I0: Intensity at center (W/m^2) *)
(* Exposure Sink/Source Term *)
m = 400/I0; (* m: Proportionality constant of intensity v.s. change \
in radical *)
d = 0.3; (* d: Proportionality constant of radical decay *)
z = 1; (* z: Axial distance from focal plane (um) *)
Intensity[x_] := I0 (W0/W[z])^2 Exp[-(2*x^2)/W[z]^2]; (* Gaussian *)
Reflection[x_] :=
I0 (W0/W[z])^2 Exp[-(2*x^2)/(400*W[z]^2)]/12 +
I0 (W0/W[z])^2 Exp[-(2*x^2)/(100*W[z]^2)]/12;
93
APPENDIX D. CODING IN MATHEMATICA 94
Exposure[x_, pos_] :=
m (Intensity[x - pos] +
Reflection[x - pos]); (* Radical generation from exposure *)
(* Probability Density of Polymerization *)
densityP[u_] := 1/(Exp[20 - 25 u] + 1);
(* Initialization *)
M = 400; dx = 20/M; (* M: # of spatial steps; dx: spacial step *)
dt = 0.002; FinalTime1 =
ExposureTime1 +
HiatusTime; (* dt: time step; FinalTime: time domain *)
alpha = .2;
cIC[x_] := 1; initc = {}; (* Initiator concentration c initialization \
*)
beta = .1;
uIC[x_] := 0; initu = {}; (* Radical concentration u initialization *)
pIC[x_, s_] :=
If[x == s, 1,
0]; initp1 = {}; initp2 = {};(* Probability Initialization of the \
first exposure *)
initPoly = {};
D.2 Initial Conditions
For[k = 0, k <= M, k++,
initu = Append[initu, uIC[k dx]];
initc = Append[initc, cIC[k dx]];
initp1 = Append[initp1, pIC[k dx, 5]];
initp2 = Append[initp2, pIC[k dx, 5 + Spacing]];
initPoly = Append[initPoly, 0];
]
u = {initu}; c = {initc}; pRight = {initp1}; pLeft = {initp1}; Poly = \
{initPoly};
APPENDIX D. CODING IN MATHEMATICA 95
D.3 First Exposure
(* First Exposure *)
For[n = 1, n*dt <= FinalTime1, n++,(* Time Domain *)
cnew = {}; unew = {}; pnewRight = {}; pnewLeft = {};
For[k = 0, k <= M, k++,(* Spatial Domain *)
Delta =
If[n*dt <= ExposureTime1, dt*Exposure[k dx, 5]*c[[n]][[k + 1]], 0];
cnew = Append[cnew,
If[0 < k && k < M,
c[[n]][[k + 1]] +
If[Poly[[n]][[k + 1]] == 0 || Poly[[n]][[k]] == 0 ||
Poly[[n]][[k + 2]] == 0, alpha, 0]*
dt*((c[[n]][[k + 2]] - 2*c[[n]][[k + 1]] +
c[[n]][[k]])/(dx^2) - c[[n]][[k + 1]] + 1) - Delta,
If[k == 0,
c[[n]][[2]] +
alpha*dt*((c[[n]][[3]] - 2*c[[n]][[2]] + c[[n]][[1]])/(dx^2) -
c[[n]][[2]] + 1) - Delta,(* Left Boundary *)
c[[n]][[M]] +
alpha*dt*((c[[n]][[M + 1]] - 2*c[[n]][[M]] +
c[[n]][[M - 1]])/(dx^2) - c[[n]][[M]] + 1) - Delta](*
Right Boundary *)
]
];
unew = Append[unew,
If[0 < k && k < M,
(1 - d*dt*(u[[n]][[k + 1]] + 1))*u[[n]][[k + 1]] +
If[Poly[[n]][[k + 1]] == 0 || Poly[[n]][[k]] == 0 ||
Poly[[n]][[k + 2]] == 0, beta, 0]*
dt*(u[[n]][[k + 2]] - 2*u[[n]][[k + 1]] +
u[[n]][[k]])/(dx^2) + 2*Delta,
If[k == 0,
u[[n]][[2]] +
beta*dt*(u[[n]][[3]] - 2*u[[n]][[2]] + u[[n]][[1]])/(dx^2) +
2*Delta,(* Left Boundary *)
u[[n]][[M]] +
beta*dt*(u[[n]][[M + 1]] - 2*u[[n]][[M]] +
u[[n]][[M - 1]])/(dx^2) + 2*Delta](* Right Boundary *)
]
];
APPENDIX D. CODING IN MATHEMATICA 96
pnewRight = Append[pnewRight,
If[0 < k && k < M,
pRight[[n]][[k + 1]]*(1 - densityP[u[[n]][[k + 1]]]) +
pRight[[n]][[k]]*densityP[u[[n]][[k]]],
If[k == 0,
pRight[[n]][[2]]*(1 - densityP[u[[n]][[2]]]) +
pRight[[n]][[1]]*densityP[u[[n]][[1]]],(* Left Boundary *)
pRight[[n]][[M]]*(1 - densityP[u[[n]][[M]]]) +
pRight[[n]][[M - 1]]*densityP[u[[n]][[M - 1]]]] (*
Right Boundary *)
]
];
pnewLeft = Append[pnewLeft,
If[0 < k && k < M,
pLeft[[n]][[k + 1]]*(1 - densityP[u[[n]][[k + 1]]]) +
pLeft[[n]][[k + 2]]*densityP[u[[n]][[k + 2]]],
If[k == 0,
pLeft[[n]][[2]]*(1 - densityP[u[[n]][[2]]]) +
pLeft[[n]][[3]]*densityP[u[[n]][[3]]],(* Left Boundary *)
pLeft[[n]][[M]]*(1 - densityP[u[[n]][[M]]]) +
pLeft[[n]][[M + 1]]*densityP[u[[n]][[M + 1]]]] (*
Right Boundary *)
]
];
];
c = Append[c, cnew]; u = Append[u, unew];
pRight = Append[pRight, pnewRight]; pLeft = Append[pLeft, pnewLeft];
(* Expected value for the boundaries in terms of box number *)
leftExpect = Sum[pnewLeft[[i]]*i, {i, 1, Length[pnewLeft]}];
rightExpect = Sum[pnewRight[[i]]*i, {i, 1, Length[pnewRight]}];
Polynew = {};
For[k = 1, k <= M + 1, k++,
Polynew =
Append[Polynew,
If[(leftExpect < k && k < rightExpect) || Poly[[-1]][[k]] == 1, 1,
0]]
];
Poly = Append[Poly, Polynew];
]
FirstDotSize = Total[Poly[[-1]]]
APPENDIX D. CODING IN MATHEMATICA 97
D.4 Second Exposure
(* Second Exposure *)
FinalTime2 = FinalTime1 + ExposureTime2 + 1;
pRight = {initp2}; pLeft = {initp2};
IndexOffSet = FinalTime1/dt;
For[n = IndexOffSet + 1, n*dt <= FinalTime2, n++,(* Time Domain *)
cnew = {}; unew = {}; pnewRight = {}; pnewLeft = {};
For[k = 0, k <= M, k++,(* Spatial Domain *)
Delta =
If[n*dt <= FinalTime1 + ExposureTime2,
dt*Exposure[k dx, 5 + Spacing]*c[[n]][[k + 1]], 0];
cnew = Append[cnew,
If[0 < k && k < M,
c[[n]][[k + 1]] +
If[Poly[[n]][[k + 1]] == 0 || Poly[[n]][[k]] == 0 ||
Poly[[n]][[k + 2]] == 0, alpha, 0]*
dt*(c[[n]][[k + 2]] - 2*c[[n]][[k + 1]] + c[[n]][[k]])/(dx^2) -
Delta,
If[k == 0,
c[[n]][[2]] +
alpha*dt*(c[[n]][[3]] - 2*c[[n]][[2]] + c[[n]][[1]])/(dx^2) -
Delta,(* Left Boundary *)
c[[n]][[M]] +
alpha*dt*(c[[n]][[M + 1]] - 2*c[[n]][[M]] +
c[[n]][[M - 1]])/(dx^2) - Delta](* Right Boundary *)
]
];
unew = Append[unew,
If[0 < k && k < M,
(1 - d*dt*(u[[n]][[k + 1]] + 1))*u[[n]][[k + 1]] +
If[Poly[[n]][[k + 1]] == 0 || Poly[[n]][[k]] == 0 ||
Poly[[n]][[k + 2]] == 0, beta, 0]*
dt*(u[[n]][[k + 2]] - 2*u[[n]][[k + 1]] +
u[[n]][[k]])/(dx^2) + 2*Delta,
If[k == 0,
u[[n]][[2]] +
beta*dt*(u[[n]][[3]] - 2*u[[n]][[2]] + u[[n]][[1]])/(dx^2) +
2*Delta,(* Left Boundary *)
u[[n]][[M]] +
beta*dt*(u[[n]][[M + 1]] - 2*u[[n]][[M]] +
APPENDIX D. CODING IN MATHEMATICA 98
u[[n]][[M - 1]])/(dx^2) + 2*Delta](* Right Boundary *)
]
];
pnewRight = Append[pnewRight,
If[0 < k && k < M,
pRight[[n - IndexOffSet]][[
k + 1]]*(1 - densityP[u[[n]][[k + 1]]]) +
pRight[[n - IndexOffSet]][[k]]*densityP[u[[n]][[k]]],
If[k == 0,
pRight[[n - IndexOffSet]][[2]]*(1 - densityP[u[[n]][[2]]]) +
pRight[[n - IndexOffSet]][[1]]*densityP[u[[n]][[1]]],(*
Left Boundary *)
pRight[[n - IndexOffSet]][[M]]*(1 - densityP[u[[n]][[M]]]) +
pRight[[n - IndexOffSet]][[M - 1]]*densityP[u[[n]][[M - 1]]]] (*
Right Boundary *)
]
];
pnewLeft = Append[pnewLeft,
If[0 < k && k < M,
pLeft[[n - IndexOffSet]][[
k + 1]]*(1 - densityP[u[[n]][[k + 1]]]) +
pLeft[[n - IndexOffSet]][[k + 2]]*densityP[u[[n]][[k + 2]]],
If[k == 0,
pLeft[[n - IndexOffSet]][[2]]*(1 - densityP[u[[n]][[2]]]) +
pLeft[[n - IndexOffSet]][[3]]*densityP[u[[n]][[3]]],(*
Left Boundary *)
pLeft[[n - IndexOffSet]][[M]]*(1 - densityP[u[[n]][[M]]]) +
pLeft[[n - IndexOffSet]][[M + 1]]*densityP[u[[n]][[M + 1]]]] (*
Right Boundary *)
]
];
];
c = Append[c, cnew]; u = Append[u, unew];
pRight = Append[pRight, pnewRight]; pLeft = Append[pLeft, pnewLeft];
(* Expected value for the boundaries in terms of box number *)
leftExpect = Sum[pnewLeft[[i]]*i, {i, 1, Length[pnewLeft]}];
rightExpect = Sum[pnewRight[[i]]*i, {i, 1, Length[pnewRight]}];
Polynew = {};
For[k = 1, k <= M + 1, k++,
Polynew =
Append[Polynew,
APPENDIX D. CODING IN MATHEMATICA 99
If[(leftExpect < k && k < rightExpect) || Poly[[-1]][[k]] == 1, 1,
0]]
];
Poly = Append[Poly, Polynew];
]
SecondDotSize = Total[Poly[[-1]]] - FirstDotSize
PercentIncrease = (SecondDotSize - FirstDotSize)/FirstDotSize // N
D.5 Showing the Result
Manipulate[
Show[
ListLinePlot[
{
Table[{20 (i - 1)/(Length[u[[n]]] - 1), u[[n]][[i]]}, {i, 1,
Length[u[[n]]]}],
Table[{20 (i - 1)/(Length[c[[n]]] - 1), c[[n]][[i]]}, {i, 1,
Length[c[[n]]]}]
},
Axes -> False, Frame -> True, PlotRange -> {{0, 20}, {0, 4/2}},
ImageSize -> 800, AspectRatio -> 1/2, PlotStyle -> Thickness[1/150],
Epilog -> {Text[
Style[ToString[(n - 1)*dt] <> "s", Bold, 48], {18, 1.8}]},
LabelStyle -> {Bold, 32}
],
ListLinePlot[
Table[{20 (i - 1)/(Length[Poly[[n]]] - 1),
Poly[[n]][[i]] + 1/2}, {i, 1, Length[Poly[[n]]]}],
PlotStyle -> Thickness[1/200],
ColorFunction -> {Function[{x, y},
If[y == 1, Black, RGBColor[0, 0, 0, 0]]]}
]
],
{n, 1, Length[c], 1}
]
Appendix E
Sample Simulation
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.1s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.2s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.3s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.4s
100
APPENDIX E. SAMPLE SIMULATION 101
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.5s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.6s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.7s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.8s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
0.9s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.1s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.2s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.3s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.4s
APPENDIX E. SAMPLE SIMULATION 102
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.5s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.6s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.7s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.8s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
1.9s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.1s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.2s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.3s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.4s
APPENDIX E. SAMPLE SIMULATION 103
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.5s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.6s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.7s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.8s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
2.9s
0
5
10
15
20
0.0
0.5
1.0
1.5
2.0
3.s
Appendix F
General Solutions to y
0
= ky(1 − y)
We first rewrite the differential equation into
dy
dx
− ky = −ky
2
. (F.1)
Let u = y
−1
, then y = u
−1
, thus
dy
du
= −u
−2
. (F.2)
By Chain Rule, it follows that
dy
dx
=
dy
du
·
du
dx
= −u
−2
·
du
dx
. (F.3)
Substitute (F.3) into (F.1), we obtain
−u
−2
·
du
dx
− ku
−1
= −ku
−2
. (F.4)
Multiply both sides by −u
2
, we obtain
du
dx
+ ku = k. (F.5)
104
APPENDIX F. GENERAL SOLUTIONS TO Y
0
= KY (1 −Y ) 105
Let the integrating factor µ(x) = e
R
kdx
= e
kx
. We multiply both sides by the integrating
factor, then integrate over x, we get
Z
e
kx
·
du
dx
+ ke
kx
· u
dx =
Z
ke
kx
dx. (F.6)
The integrand of the left-hand side is the derivative of e
kx
·u(x), and the integrand of the
right-hand side is the derivative of e
kx
. Thus we have
e
kx
· u(x) = e
kx
+ C. (F.7)
We solve for u(x), and get
u(x) = 1 + e
c−kx
. (F.8)
Thus the general solution to y(x), in closed form, is
y(x) =
1
1 + e
c−kx
, (F.9)
where c and k are constants, and k > 0.
Bibliography
[Abe62] I. D. Abella. Optical double-photon absorption in cesium vapor. Phys. Rev.
Lett., 9(11):453–455, Dec 1962.
[Air35] G. B. Airy. On the diffraction of an object-glass with circular aperture. Trans-
actions of the Cambridge Philosophical Society, 5:283–291, 1835.
[Far82] Stanley J. Farlow. Partial Differential Equations for Scientists and Engineers.
Dover Publications, Inc., 1982.
[FW13] Joachim Fischer and Martin Wegener. Three-dimensional optical laser lithog-
raphy beyond the diffraction limit. Laser and Photonics Reviews, 7(1):22–44,
2013.
[IKT
+
08] Kazuyoshi Itoga, Jun Kobayashi, Yukiko Tsuda, Masayuki Yamato, and Teruo
Okano. Second-generation maskless photolithography device for surface mi-
cropatterning and microfluidic channel fabrication. Analytical Chemistry,
80(4):1323–1327, 2008. PMID: 18211096.
[Jae02] Richard C. Jaeger. Introduction to Microelectronic Fabrication, volume 5 of
Modular Series on Solid State Devices, chapter 2 Lithography, pages 17–42.
Prentice Hall, 2nd edition, 2002.
[KOM
+
13] Michelle L. Kovarik, Douglas M. Ornoff, Adam T. Melvin, Nicholas C. Dobes,
Yuli Wang, Alexandra J. Dickinson, Philip C. Gach, Pavak K. Shah, and
Nancy L. Allbritton. Micro total analysis systems: Fundamental advances
and applications in the laboratory, clinic, and field. Analytical Chemistry,
85(2):451–472, 2013.
[Moo65] Gorden E. Moore. Cramming more components onto integrated circuits. Elec-
tronics, 1965.
106
BIBLIOGRAPHY 107
[PY05] Michael F. Perry and Gerald W. Young. A mathematical model for photopoly-
merization from a stationary laser light source. Macromolecular Theory and
Simulations, 14(1):26–39, 2005.
[RC97] P. Rai-Choudhury, editor. Handbook of Microlithography, Micromachining,
and Microfabrication, volume 1, chapter 2 Electron beam lithography, pages
139–250. Society of Photo-Optical Instrumentation Engineers, March 1997.
[Rev84] D. Revus. Markov Chains. Elsevier, 1984.
[Rub74] Robert Rubinstein. Examples of nonsolvable partial differential equations.
Trans. Amer. Math. Soc., 199:123–129, 1974.
[Sig13] Sigma-Aldrich. Diethyl ether, Material Safety Data Sheet, 5.2 edition, August
2013.
[ST91] Bahaa E. Saleh and Malvin Carl Teich. Fundamentals of Photonics. Wiley
series in pure and applied optics. Wiley, 1991.
[Sza10] Csilla Szabo. Markov Models for Actin Polymer Dynamics and Cell Membrane
Protrusion. PhD thesis, Rensselaer Polytechnic Institute, May 2010.
[TBP
+
11] Giovanna Tomaiuolo, Mario Barra, Valentina Preziosi, Antonio Cassinese,
Bruno Rotoli, and Stefano Guido. Microfluidics analysis of red blood cell
membrane viscoelasticity. Lab Chip, 11:449–454, 2011.
[Tho95] J. W. Thomas. Numerical Differential Equations: Finite Difference Methods.
Springer, 1995.
[ZNWS13] Yi Zheng, John Nguyen, Chen Wang, and Yu Sun. Electrical measurement of
red blood cell deformability on a microfluidic device. Lab Chip, 13:3275–3283,
2013.
[ZRM
+
03] Dong S. Zhao, Binayak Roy, Matthew T. McCormick, Werner G. Kuhr, and
Sara A. Brazill. Rapid fabrication of a poly(dimethylsiloxane) microfluidic
capillary gel electrophoresis system utilizing high precision machining. Lab
Chip, 3:93–99, 2003.
